Fall 2024 GRASP SFI - Baxi Chong, Georgia Institute of Technology
Baxi Chong, a researcher at Georgia Institute of Technology, discusses how mechanical intelligence and redundant robot designs can improve locomotion in noisy, complex terrain, and how to balance that with sensing and control.
In this talk
- 0:02
Why robots need mechanical intelligence
Chong compares quadruped robots with animals that handle difficult terrain, heavy loads, and recovery from falls. He argues that robots relying on sensing and computation can struggle to generalize across environments and behaviors. Smaller robots face a particular challenge because they must move through clutter while interacting with several objects at once.
- 12:39
Redundant legs improve reliability
Chong frames locomotion in unpredictable terrain as a noisy-channel problem. In simulations and robot trials, adding legs bounded the variation in arrival time, while extra time alone could ensure eventual completion without making arrival predictable. In tests across complex terrain, robots with more legs got stuck less often.
- 26:40
Using shape changes to gain speed
To improve the speed of leggy robots, Chong explores how coordinated body and leg motion can generate propulsion. The best combination depends on the number of legs: a 16-legged design favored one motion pattern, while an eight-legged design performed best with a blend. This creates a tradeoff between agility and robustness, which lightweight contact sensors and simple feedback can help manage.
- 43:13
Designing simpler robots from contact patterns
Chong models a robot’s possible contact patterns as a graph, with contact changes linked to movement. Searching that space produced an asymmetric gait that could run at a higher tempo, and revealed motors that stayed inactive throughout the gait. Removing those motors could reduce weight while preserving the robot’s forward movement.
- 7:40
Where the approach could matter
Chong points to inspection inside collapsed buildings, precision agriculture, and movement over clutter in homes as potential uses for small, capable robots. A lighter robot could inspect unstable rubble with less force, while agricultural robots could navigate varied terrain to examine individual crops. He also argues that mechanical redundancy can provide robustness when sensors are unavailable or unreliable.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.