Quadrupedal Footstep Planning using Learned Motion Models of a Black-Box Controller

Ilyass Taouil,Giulio Turrisi,Daniel Schleich,Victor Barasuol,Claudio Semini,Sven Behnke

2023-07-23

Abstract:Legged robots are increasingly entering new domains and applications, including search and rescue, inspection, and logistics. However, for such systems to be valuable in real-world scenarios, they must be able to autonomously and robustly navigate irregular terrains. In many cases, robots that are sold on the market do not provide such abilities, being able to perform only blind locomotion. Furthermore, their controller cannot be easily modified by the end-user, requiring a new and time-consuming control synthesis. In this work, we present a fast local motion planning pipeline that extends the capabilities of a black-box walking controller that is only able to track high-level reference velocities. More precisely, we learn a set of motion models for such a controller that maps high-level velocity commands to Center of Mass (CoM) and footstep motions. We then integrate these models with a variant of the A star algorithm to plan the CoM trajectory, footstep sequences, and corresponding high-level velocity commands based on visual information, allowing the quadruped to safely traverse irregular terrains at demand.

Robotics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to enable legged robots (especially quadruped robots) to navigate autonomously and robustly in unknown and irregular terrains. Specifically, many robots on the market can only perform blind locomotion, that is, their controllers can only track high - level velocity commands, but cannot receive target footstep positions or adjust their movements according to visual information. This makes these robots prone to failure in complex environments.

To overcome this limitation, the authors propose a fast local motion - planning pipeline, which extends the capabilities of the black - box walking controller, enabling it to plan the center - of - mass (CoM) trajectory, footstep sequence, and corresponding high - level velocity commands based on visual information. Through this method, quadruped robots can safely traverse irregular terrains.

### Main problems
1. **Limitations of existing controllers**: The controllers of many quadruped robots on the market can only track high - level velocity commands and cannot receive specific footstep position commands, resulting in their poor performance in complex environments.
2. **Lack of perception ability**: These robots lack the ability to perceive the surrounding environment and adjust their movements accordingly, especially on irregular terrains.
3. **Motion planning that needs improvement**: In order to enable robots to navigate autonomously in complex environments, a method that can combine visual information for motion planning is required.

### Solutions
The authors propose a learning - based motion - model framework, which can predict the behavior of the black - box controller and combine it with an A* algorithm variant to plan the optimal footstep sequence and velocity commands. Specific steps include:
- **Learning the motion model**: By learning the behavior of the black - box controller, predict the CoM displacement and footstep positions under given velocity commands.
- **Terrain mapping**: Use a depth camera to reconstruct the local terrain and generate a foot - cost map, identifying safe and unsafe footstep positions.
- **Motion planning**: Utilize an A* algorithm variant, combine the above - mentioned model and terrain information, and plan the optimal footstep sequence and velocity commands.

Through this method, robots can enhance their perception and autonomous - navigation abilities while maintaining the original controller, thus better adapting to complex environments.