#robotics101 search results

@mvpsd Girls are victorious once again after an epic 10 minute battle in the arena! #awesomejobladies #robotics101

TomMilbrandt's tweet image. @mvpsd Girls are victorious once again after an epic 10 minute battle in the arena! #awesomejobladies #robotics101

GTT took a trip to @NASA_Johnson to check out demonstrations w simulators & interview astronauts! #robotics101 #NASA

GoToTeam's tweet image. GTT took a trip to @NASA_Johnson to check out demonstrations w simulators & interview astronauts! #robotics101 #NASA

The crowd gathers to see the plinko project finale #robotics101 @STEMcha @STEMpythons

STEMFabLab's tweet image. The crowd gathers to see the plinko project finale #robotics101 @STEMcha @STEMpythons
STEMFabLab's tweet image. The crowd gathers to see the plinko project finale #robotics101 @STEMcha @STEMpythons

Some final tweaks (including pneumatic models) to finish off our mock strategy workshop and officially complete #robotics101! We can’t wait for #FIRSTPowerUp and the new year! @STEMcha @STEMFabLab @hamcoschools


Having our second #Robotics101 meeting today. Project Plinko is in full swing with the start of prototyping today of different game boards.

STEMpythons's tweet image. Having our second #Robotics101 meeting today. Project Plinko is in full swing with the start of prototyping today of different game boards.
STEMpythons's tweet image. Having our second #Robotics101 meeting today. Project Plinko is in full swing with the start of prototyping today of different game boards.
STEMpythons's tweet image. Having our second #Robotics101 meeting today. Project Plinko is in full swing with the start of prototyping today of different game boards.
STEMpythons's tweet image. Having our second #Robotics101 meeting today. Project Plinko is in full swing with the start of prototyping today of different game boards.

#BigBytes a startup at @NICLahore conducting a #Robotics101 workshop. Snap shows lead Founder talking about #Aurdinos

fsherjan's tweet image. #BigBytes a startup at @NICLahore conducting a #Robotics101 workshop. Snap shows lead Founder talking about #Aurdinos
fsherjan's tweet image. #BigBytes a startup at @NICLahore conducting a #Robotics101 workshop. Snap shows lead Founder talking about #Aurdinos

A primer on event cameras by @davsca1: definitely the biggest advocate of event cameras in the robotics and computer vision world #robotics101 #cvpr2025

abursuc's tweet image. A primer on event cameras by @davsca1: definitely the biggest advocate of event cameras in the robotics and computer vision world 
#robotics101 #cvpr2025
abursuc's tweet image. A primer on event cameras by @davsca1: definitely the biggest advocate of event cameras in the robotics and computer vision world 
#robotics101 #cvpr2025
abursuc's tweet image. A primer on event cameras by @davsca1: definitely the biggest advocate of event cameras in the robotics and computer vision world 
#robotics101 #cvpr2025
abursuc's tweet image. A primer on event cameras by @davsca1: definitely the biggest advocate of event cameras in the robotics and computer vision world 
#robotics101 #cvpr2025

DreamDrive: combining smartly scenario generation and reconstruction putting together lots of cool prior works (EmerNeRF, STORM) and foundation models. Learning to drive in a generative 4D world @iamborisi #robotics101 #cvpr2025

abursuc's tweet image. DreamDrive: combining smartly scenario generation and reconstruction putting together lots of cool prior works (EmerNeRF, STORM) and foundation models.
Learning to drive in a generative 4D world @iamborisi 
#robotics101 #cvpr2025

I'm sure that everybody at the #robotics101 tutorial by @davsca1 now would like to be the Marvin kid playing as expert for drone racing #cvpr2025

abursuc's tweet image. I'm sure that everybody at the #robotics101 tutorial by @davsca1 now would like to be the Marvin kid playing as expert for drone racing #cvpr2025

NavSimV2: finding a middle ground in the closed-loop/open-loop dichotomy -> pseudo-closed-loop evaluation. Pre-generate sensor rendering, generate multiple rollouts and select likeliest ones. Approaches closed-loop by a lot. @iamborisi #robotics101 #cvpr2025

abursuc's tweet image. NavSimV2: finding a middle ground in the closed-loop/open-loop dichotomy -> pseudo-closed-loop evaluation.
Pre-generate sensor rendering, generate multiple rollouts and select likeliest ones. 
Approaches closed-loop by a lot.
@iamborisi 
#robotics101 #cvpr2025
abursuc's tweet image. NavSimV2: finding a middle ground in the closed-loop/open-loop dichotomy -> pseudo-closed-loop evaluation.
Pre-generate sensor rendering, generate multiple rollouts and select likeliest ones. 
Approaches closed-loop by a lot.
@iamborisi 
#robotics101 #cvpr2025

A strong motivation for working on agility by @davsca1 Multi-rotor drone battery is limited to 30 min. If you fly faster, the drones can be actually more productive #cvpr2024 #robotics101

abursuc's tweet image. A strong motivation for working on agility by @davsca1 
Multi-rotor drone battery is limited to 30 min. If you fly faster, the drones can be actually more productive #cvpr2024 #robotics101

There is still a domain gap that can be addressed with a few tricks: - fine-tuning on prev. real rollouts - use directly RL actions instead of plan+control. RL is better: it directly optimizes a task-level objective & leverage domain randomization @davsca1 #robotics101 #cvpr2025

abursuc's tweet image. There is still a domain gap that can be addressed with a few tricks:
- fine-tuning on prev. real rollouts
- use directly RL actions instead of plan+control. RL is better: it directly optimizes a task-level objective & leverage domain randomization @davsca1 #robotics101 #cvpr2025
abursuc's tweet image. There is still a domain gap that can be addressed with a few tricks:
- fine-tuning on prev. real rollouts
- use directly RL actions instead of plan+control. RL is better: it directly optimizes a task-level objective & leverage domain randomization @davsca1 #robotics101 #cvpr2025
abursuc's tweet image. There is still a domain gap that can be addressed with a few tricks:
- fine-tuning on prev. real rollouts
- use directly RL actions instead of plan+control. RL is better: it directly optimizes a task-level objective & leverage domain randomization @davsca1 #robotics101 #cvpr2025

The magic of RL when done and applied right: training a drone racing policy in simulation took 16 minutes with not SoTA hardware: learning to fly in 10% of steps and to race in the remaining 90% steps @davsca1 #robotics101 #cvpr2025

abursuc's tweet image. The magic of RL when done and applied right: training a drone racing policy in simulation took 16 minutes with not SoTA hardware: learning to fly in 10% of steps and to race in the remaining 90% steps
@davsca1 #robotics101 #cvpr2025
abursuc's tweet image. The magic of RL when done and applied right: training a drone racing policy in simulation took 16 minutes with not SoTA hardware: learning to fly in 10% of steps and to race in the remaining 90% steps
@davsca1 #robotics101 #cvpr2025
abursuc's tweet image. The magic of RL when done and applied right: training a drone racing policy in simulation took 16 minutes with not SoTA hardware: learning to fly in 10% of steps and to race in the remaining 90% steps
@davsca1 #robotics101 #cvpr2025
abursuc's tweet image. The magic of RL when done and applied right: training a drone racing policy in simulation took 16 minutes with not SoTA hardware: learning to fly in 10% of steps and to race in the remaining 90% steps
@davsca1 #robotics101 #cvpr2025

A recipe from the domain generalization trenches for drone navigation by @davsca1: - training on synthetic images doesn't work at all - however if you train on abstract representations (optical flow, depth maps, etc.) you're doing way way better #robotics101 #cvpr2025

abursuc's tweet image. A recipe from the domain generalization trenches for drone navigation by @davsca1: 
- training on synthetic images doesn't work at all
- however if you train on abstract representations (optical flow, depth maps, etc.) you're doing way way better

#robotics101 #cvpr2025

The magic effectiveness of end-to-end driving w/ a Vision-Language-Action model w/ auto-regressive backbones: leverage pretrained foundation models (even generic ones work ~well) and finetune for motion planning with minimal supervision @iamborisi #robotics101 #cvpr2025

abursuc's tweet image. The magic effectiveness of end-to-end driving w/ a Vision-Language-Action model w/ auto-regressive backbones: leverage pretrained foundation models (even generic ones work ~well) and finetune for motion planning with minimal supervision
@iamborisi 
#robotics101 #cvpr2025
abursuc's tweet image. The magic effectiveness of end-to-end driving w/ a Vision-Language-Action model w/ auto-regressive backbones: leverage pretrained foundation models (even generic ones work ~well) and finetune for motion planning with minimal supervision
@iamborisi 
#robotics101 #cvpr2025
abursuc's tweet image. The magic effectiveness of end-to-end driving w/ a Vision-Language-Action model w/ auto-regressive backbones: leverage pretrained foundation models (even generic ones work ~well) and finetune for motion planning with minimal supervision
@iamborisi 
#robotics101 #cvpr2025
abursuc's tweet image. The magic effectiveness of end-to-end driving w/ a Vision-Language-Action model w/ auto-regressive backbones: leverage pretrained foundation models (even generic ones work ~well) and finetune for motion planning with minimal supervision
@iamborisi 
#robotics101 #cvpr2025

Robots don’t think for themselves (yet 😏), but they’re changing the way we live, work, and play. #TechExplained #Robotics101 #MicroLearning #AIandRobots #STEMEducation #SmartTech #HowStuffWorks


Smart idea for tokenization in autoregressive AV models: instead of tokenizing images, project them in BEV volumes & tokenize voxels with triplane decomposition. It works even when removing/adding cameras while keeping the token number constant @iamborisi #robotics101 #cvpr2025

abursuc's tweet image. Smart idea for tokenization in autoregressive AV models: instead of tokenizing images, project them in BEV volumes & tokenize voxels with triplane decomposition. It works even when removing/adding cameras while keeping the token number constant
@iamborisi 
#robotics101 #cvpr2025
abursuc's tweet image. Smart idea for tokenization in autoregressive AV models: instead of tokenizing images, project them in BEV volumes & tokenize voxels with triplane decomposition. It works even when removing/adding cameras while keeping the token number constant
@iamborisi 
#robotics101 #cvpr2025
abursuc's tweet image. Smart idea for tokenization in autoregressive AV models: instead of tokenizing images, project them in BEV volumes & tokenize voxels with triplane decomposition. It works even when removing/adding cameras while keeping the token number constant
@iamborisi 
#robotics101 #cvpr2025
abursuc's tweet image. Smart idea for tokenization in autoregressive AV models: instead of tokenizing images, project them in BEV volumes & tokenize voxels with triplane decomposition. It works even when removing/adding cameras while keeping the token number constant
@iamborisi 
#robotics101 #cvpr2025

The magic effectiveness of end-to-end driving w/ a Vision-Language-Action model w/ auto-regressive backbones: leverage pretrained foundation models (even generic ones work ~well) and finetune for motion planning with minimal supervision @iamborisi #robotics101 #cvpr2025

abursuc's tweet image. The magic effectiveness of end-to-end driving w/ a Vision-Language-Action model w/ auto-regressive backbones: leverage pretrained foundation models (even generic ones work ~well) and finetune for motion planning with minimal supervision
@iamborisi 
#robotics101 #cvpr2025
abursuc's tweet image. The magic effectiveness of end-to-end driving w/ a Vision-Language-Action model w/ auto-regressive backbones: leverage pretrained foundation models (even generic ones work ~well) and finetune for motion planning with minimal supervision
@iamborisi 
#robotics101 #cvpr2025
abursuc's tweet image. The magic effectiveness of end-to-end driving w/ a Vision-Language-Action model w/ auto-regressive backbones: leverage pretrained foundation models (even generic ones work ~well) and finetune for motion planning with minimal supervision
@iamborisi 
#robotics101 #cvpr2025
abursuc's tweet image. The magic effectiveness of end-to-end driving w/ a Vision-Language-Action model w/ auto-regressive backbones: leverage pretrained foundation models (even generic ones work ~well) and finetune for motion planning with minimal supervision
@iamborisi 
#robotics101 #cvpr2025

NavSimV2: finding a middle ground in the closed-loop/open-loop dichotomy -> pseudo-closed-loop evaluation. Pre-generate sensor rendering, generate multiple rollouts and select likeliest ones. Approaches closed-loop by a lot. @iamborisi #robotics101 #cvpr2025

abursuc's tweet image. NavSimV2: finding a middle ground in the closed-loop/open-loop dichotomy -> pseudo-closed-loop evaluation.
Pre-generate sensor rendering, generate multiple rollouts and select likeliest ones. 
Approaches closed-loop by a lot.
@iamborisi 
#robotics101 #cvpr2025
abursuc's tweet image. NavSimV2: finding a middle ground in the closed-loop/open-loop dichotomy -> pseudo-closed-loop evaluation.
Pre-generate sensor rendering, generate multiple rollouts and select likeliest ones. 
Approaches closed-loop by a lot.
@iamborisi 
#robotics101 #cvpr2025

DreamDrive: combining smartly scenario generation and reconstruction putting together lots of cool prior works (EmerNeRF, STORM) and foundation models. Learning to drive in a generative 4D world @iamborisi #robotics101 #cvpr2025

abursuc's tweet image. DreamDrive: combining smartly scenario generation and reconstruction putting together lots of cool prior works (EmerNeRF, STORM) and foundation models.
Learning to drive in a generative 4D world @iamborisi 
#robotics101 #cvpr2025

And now @iamborisi takes the stage with Developing, Validating, and Safely Deploying Physical AI in Autonomous Vehicles #robotics101 #cvpr2025

abursuc's tweet image. And now @iamborisi takes the stage with Developing, Validating, and Safely Deploying Physical AI in Autonomous Vehicles

#robotics101 #cvpr2025

A primer on event cameras by @davsca1: definitely the biggest advocate of event cameras in the robotics and computer vision world #robotics101 #cvpr2025

abursuc's tweet image. A primer on event cameras by @davsca1: definitely the biggest advocate of event cameras in the robotics and computer vision world 
#robotics101 #cvpr2025
abursuc's tweet image. A primer on event cameras by @davsca1: definitely the biggest advocate of event cameras in the robotics and computer vision world 
#robotics101 #cvpr2025
abursuc's tweet image. A primer on event cameras by @davsca1: definitely the biggest advocate of event cameras in the robotics and computer vision world 
#robotics101 #cvpr2025
abursuc's tweet image. A primer on event cameras by @davsca1: definitely the biggest advocate of event cameras in the robotics and computer vision world 
#robotics101 #cvpr2025

There is still a domain gap that can be addressed with a few tricks: - fine-tuning on prev. real rollouts - use directly RL actions instead of plan+control. RL is better: it directly optimizes a task-level objective & leverage domain randomization @davsca1 #robotics101 #cvpr2025

abursuc's tweet image. There is still a domain gap that can be addressed with a few tricks:
- fine-tuning on prev. real rollouts
- use directly RL actions instead of plan+control. RL is better: it directly optimizes a task-level objective & leverage domain randomization @davsca1 #robotics101 #cvpr2025
abursuc's tweet image. There is still a domain gap that can be addressed with a few tricks:
- fine-tuning on prev. real rollouts
- use directly RL actions instead of plan+control. RL is better: it directly optimizes a task-level objective & leverage domain randomization @davsca1 #robotics101 #cvpr2025
abursuc's tweet image. There is still a domain gap that can be addressed with a few tricks:
- fine-tuning on prev. real rollouts
- use directly RL actions instead of plan+control. RL is better: it directly optimizes a task-level objective & leverage domain randomization @davsca1 #robotics101 #cvpr2025

The magic of RL when done and applied right: training a drone racing policy in simulation took 16 minutes with not SoTA hardware: learning to fly in 10% of steps and to race in the remaining 90% steps @davsca1 #robotics101 #cvpr2025

abursuc's tweet image. The magic of RL when done and applied right: training a drone racing policy in simulation took 16 minutes with not SoTA hardware: learning to fly in 10% of steps and to race in the remaining 90% steps
@davsca1 #robotics101 #cvpr2025
abursuc's tweet image. The magic of RL when done and applied right: training a drone racing policy in simulation took 16 minutes with not SoTA hardware: learning to fly in 10% of steps and to race in the remaining 90% steps
@davsca1 #robotics101 #cvpr2025
abursuc's tweet image. The magic of RL when done and applied right: training a drone racing policy in simulation took 16 minutes with not SoTA hardware: learning to fly in 10% of steps and to race in the remaining 90% steps
@davsca1 #robotics101 #cvpr2025
abursuc's tweet image. The magic of RL when done and applied right: training a drone racing policy in simulation took 16 minutes with not SoTA hardware: learning to fly in 10% of steps and to race in the remaining 90% steps
@davsca1 #robotics101 #cvpr2025

A recipe from the domain generalization trenches for drone navigation by @davsca1: - training on synthetic images doesn't work at all - however if you train on abstract representations (optical flow, depth maps, etc.) you're doing way way better #robotics101 #cvpr2025

abursuc's tweet image. A recipe from the domain generalization trenches for drone navigation by @davsca1: 
- training on synthetic images doesn't work at all
- however if you train on abstract representations (optical flow, depth maps, etc.) you're doing way way better

#robotics101 #cvpr2025

I'm sure that everybody at the #robotics101 tutorial by @davsca1 now would like to be the Marvin kid playing as expert for drone racing #cvpr2025

abursuc's tweet image. I'm sure that everybody at the #robotics101 tutorial by @davsca1 now would like to be the Marvin kid playing as expert for drone racing #cvpr2025

A strong motivation for working on agility by @davsca1 Multi-rotor drone battery is limited to 30 min. If you fly faster, the drones can be actually more productive #cvpr2024 #robotics101

abursuc's tweet image. A strong motivation for working on agility by @davsca1 
Multi-rotor drone battery is limited to 30 min. If you fly faster, the drones can be actually more productive #cvpr2024 #robotics101

#Robotics101 is building and engineering their @LEGO_Group robots! Lots of problem solving and teamwork is happening in Room 207 @KEllis2023

Gifted_Mrs_Lee's tweet image. #Robotics101 is building and engineering their @LEGO_Group robots! Lots of problem solving and teamwork is happening in Room 207 @KEllis2023
Gifted_Mrs_Lee's tweet image. #Robotics101 is building and engineering their @LEGO_Group robots! Lots of problem solving and teamwork is happening in Room 207 @KEllis2023
Gifted_Mrs_Lee's tweet image. #Robotics101 is building and engineering their @LEGO_Group robots! Lots of problem solving and teamwork is happening in Room 207 @KEllis2023

Are you ready to dive into the world of robotics? 🤖 Join us for an exciting session with George Okorafor as he unravels the mysteries of Robotics. To join the meeting; bit.ly/rain_inn_events #RAIN #RAIN_INN #Robotics101 #TechTalk #Innovation #FutureofTech

FlourishOlaiya's tweet image. Are you ready to dive into the world of robotics? 🤖

Join us for an exciting session with George Okorafor as he unravels the mysteries of Robotics.

To join the meeting; bit.ly/rain_inn_events

#RAIN #RAIN_INN #Robotics101 #TechTalk #Innovation #FutureofTech

New to robotics? Let's dive into the basics! Robotics LLC explores the fascinating world where science, engineering, and technology converge to create machines that mimic human actions. 🤖 #Robotics101 #EngineeringInnovation #TechTrends #RoboticsLLC #LearnRobotics

robotics_llc's tweet image. New to robotics? 

Let's dive into the basics! Robotics LLC explores the fascinating world where science, engineering, and technology converge to create machines that mimic human actions. 🤖

 #Robotics101 #EngineeringInnovation #TechTrends #RoboticsLLC #LearnRobotics

The crowd gathers to see the plinko project finale #robotics101 @STEMcha @STEMpythons

STEMFabLab's tweet image. The crowd gathers to see the plinko project finale #robotics101 @STEMcha @STEMpythons
STEMFabLab's tweet image. The crowd gathers to see the plinko project finale #robotics101 @STEMcha @STEMpythons

@mvpsd Girls are victorious once again after an epic 10 minute battle in the arena! #awesomejobladies #robotics101

TomMilbrandt's tweet image. @mvpsd Girls are victorious once again after an epic 10 minute battle in the arena! #awesomejobladies #robotics101

GTT took a trip to @NASA_Johnson to check out demonstrations w simulators & interview astronauts! #robotics101 #NASA

GoToTeam's tweet image. GTT took a trip to @NASA_Johnson to check out demonstrations w simulators & interview astronauts! #robotics101 #NASA

Having our second #Robotics101 meeting today. Project Plinko is in full swing with the start of prototyping today of different game boards.

STEMpythons's tweet image. Having our second #Robotics101 meeting today. Project Plinko is in full swing with the start of prototyping today of different game boards.
STEMpythons's tweet image. Having our second #Robotics101 meeting today. Project Plinko is in full swing with the start of prototyping today of different game boards.
STEMpythons's tweet image. Having our second #Robotics101 meeting today. Project Plinko is in full swing with the start of prototyping today of different game boards.
STEMpythons's tweet image. Having our second #Robotics101 meeting today. Project Plinko is in full swing with the start of prototyping today of different game boards.

#BigBytes a startup at @NICLahore conducting a #Robotics101 workshop. Snap shows lead Founder talking about #Aurdinos

fsherjan's tweet image. #BigBytes a startup at @NICLahore conducting a #Robotics101 workshop. Snap shows lead Founder talking about #Aurdinos
fsherjan's tweet image. #BigBytes a startup at @NICLahore conducting a #Robotics101 workshop. Snap shows lead Founder talking about #Aurdinos

A primer on event cameras by @davsca1: definitely the biggest advocate of event cameras in the robotics and computer vision world #robotics101 #cvpr2025

abursuc's tweet image. A primer on event cameras by @davsca1: definitely the biggest advocate of event cameras in the robotics and computer vision world 
#robotics101 #cvpr2025
abursuc's tweet image. A primer on event cameras by @davsca1: definitely the biggest advocate of event cameras in the robotics and computer vision world 
#robotics101 #cvpr2025
abursuc's tweet image. A primer on event cameras by @davsca1: definitely the biggest advocate of event cameras in the robotics and computer vision world 
#robotics101 #cvpr2025
abursuc's tweet image. A primer on event cameras by @davsca1: definitely the biggest advocate of event cameras in the robotics and computer vision world 
#robotics101 #cvpr2025

Join FabLab next week for Robotics 101! This class will allow your inner engineer and code master to shine as you build a motorized robot and learn to code types of robots! This class also counts as a prerequisite for future FabLab Design Days. #FabLabWorkshops #Robotics101

sciencespectrum's tweet image. Join FabLab next week for Robotics 101! This class will allow your inner engineer and code master to shine as you build a motorized robot and learn to code types of robots! This class also counts as a prerequisite for future FabLab Design Days. #FabLabWorkshops #Robotics101

The latest in #Robotics101 are #Sensors that allow #Robots to see, hear, touch and move. Vision, audio, movement and touch sensors have enabled robots to do incredible things! What does this mean for our future? #Robotics #Innovation possibility.teledyneimaging.com/robotics-101-s…

TeledyneDALSA's tweet image. The latest in #Robotics101 are #Sensors that allow #Robots to see, hear, touch and move. Vision, audio, movement and touch sensors have enabled robots to do incredible things! What does this mean for our future?  #Robotics #Innovation possibility.teledyneimaging.com/robotics-101-s…

I'm sure that everybody at the #robotics101 tutorial by @davsca1 now would like to be the Marvin kid playing as expert for drone racing #cvpr2025

abursuc's tweet image. I'm sure that everybody at the #robotics101 tutorial by @davsca1 now would like to be the Marvin kid playing as expert for drone racing #cvpr2025

DreamDrive: combining smartly scenario generation and reconstruction putting together lots of cool prior works (EmerNeRF, STORM) and foundation models. Learning to drive in a generative 4D world @iamborisi #robotics101 #cvpr2025

abursuc's tweet image. DreamDrive: combining smartly scenario generation and reconstruction putting together lots of cool prior works (EmerNeRF, STORM) and foundation models.
Learning to drive in a generative 4D world @iamborisi 
#robotics101 #cvpr2025

NavSimV2: finding a middle ground in the closed-loop/open-loop dichotomy -> pseudo-closed-loop evaluation. Pre-generate sensor rendering, generate multiple rollouts and select likeliest ones. Approaches closed-loop by a lot. @iamborisi #robotics101 #cvpr2025

abursuc's tweet image. NavSimV2: finding a middle ground in the closed-loop/open-loop dichotomy -> pseudo-closed-loop evaluation.
Pre-generate sensor rendering, generate multiple rollouts and select likeliest ones. 
Approaches closed-loop by a lot.
@iamborisi 
#robotics101 #cvpr2025
abursuc's tweet image. NavSimV2: finding a middle ground in the closed-loop/open-loop dichotomy -> pseudo-closed-loop evaluation.
Pre-generate sensor rendering, generate multiple rollouts and select likeliest ones. 
Approaches closed-loop by a lot.
@iamborisi 
#robotics101 #cvpr2025

A strong motivation for working on agility by @davsca1 Multi-rotor drone battery is limited to 30 min. If you fly faster, the drones can be actually more productive #cvpr2024 #robotics101

abursuc's tweet image. A strong motivation for working on agility by @davsca1 
Multi-rotor drone battery is limited to 30 min. If you fly faster, the drones can be actually more productive #cvpr2024 #robotics101

And now @iamborisi takes the stage with Developing, Validating, and Safely Deploying Physical AI in Autonomous Vehicles #robotics101 #cvpr2025

abursuc's tweet image. And now @iamborisi takes the stage with Developing, Validating, and Safely Deploying Physical AI in Autonomous Vehicles

#robotics101 #cvpr2025

The magic of RL when done and applied right: training a drone racing policy in simulation took 16 minutes with not SoTA hardware: learning to fly in 10% of steps and to race in the remaining 90% steps @davsca1 #robotics101 #cvpr2025

abursuc's tweet image. The magic of RL when done and applied right: training a drone racing policy in simulation took 16 minutes with not SoTA hardware: learning to fly in 10% of steps and to race in the remaining 90% steps
@davsca1 #robotics101 #cvpr2025
abursuc's tweet image. The magic of RL when done and applied right: training a drone racing policy in simulation took 16 minutes with not SoTA hardware: learning to fly in 10% of steps and to race in the remaining 90% steps
@davsca1 #robotics101 #cvpr2025
abursuc's tweet image. The magic of RL when done and applied right: training a drone racing policy in simulation took 16 minutes with not SoTA hardware: learning to fly in 10% of steps and to race in the remaining 90% steps
@davsca1 #robotics101 #cvpr2025
abursuc's tweet image. The magic of RL when done and applied right: training a drone racing policy in simulation took 16 minutes with not SoTA hardware: learning to fly in 10% of steps and to race in the remaining 90% steps
@davsca1 #robotics101 #cvpr2025

There is still a domain gap that can be addressed with a few tricks: - fine-tuning on prev. real rollouts - use directly RL actions instead of plan+control. RL is better: it directly optimizes a task-level objective & leverage domain randomization @davsca1 #robotics101 #cvpr2025

abursuc's tweet image. There is still a domain gap that can be addressed with a few tricks:
- fine-tuning on prev. real rollouts
- use directly RL actions instead of plan+control. RL is better: it directly optimizes a task-level objective & leverage domain randomization @davsca1 #robotics101 #cvpr2025
abursuc's tweet image. There is still a domain gap that can be addressed with a few tricks:
- fine-tuning on prev. real rollouts
- use directly RL actions instead of plan+control. RL is better: it directly optimizes a task-level objective & leverage domain randomization @davsca1 #robotics101 #cvpr2025
abursuc's tweet image. There is still a domain gap that can be addressed with a few tricks:
- fine-tuning on prev. real rollouts
- use directly RL actions instead of plan+control. RL is better: it directly optimizes a task-level objective & leverage domain randomization @davsca1 #robotics101 #cvpr2025

Loading...

Something went wrong.


Something went wrong.


United States Trends