Wallbounce : Push wall to navigate with Contact-Implicit MPC

TL;DR

This work uses a bi-level contact-implicit planner and hybrid model predictive controller to draft and execute a motion plan that enables highly maneuverable locomotion using non-periodic contacts to increase the control authority of unique robot morpohologies without additional hardware.

Abstract

In this work, we introduce a framework that enables highly maneuverable locomotion using non-periodic contacts. This task is challenging for traditional optimization and planning methods to handle due to difficulties in specifying contact mode sequences in real-time. To address this, we use a bi-level contact-implicit planner and hybrid model predictive controller to draft and execute a motion plan. We investigate how this method allows us to plan arm contact events on the shmoobot, a smaller ballbot, which uses an inverse mouse-ball drive to achieve dynamic balancing with a low number of actuators. Through multiple experiments we show how the arms allow for acceleration, deceleration and dynamic obstacle avoidance that are not achievable with the mouse-ball drive alone. This demonstrates how a holistic approach to locomotion can increase the control authority of unique robot morpohologies without additional hardware by leveraging robot arms that are typically used only for manipulation. Project website: https://cmushmoobot.github.io/Wallbounce

Figures (7)

• Figure 1: Time-lapse picture of CMU shmoobot making a sharp turn by pushing wall.
• Figure 2: (a) CMU shmoobot. (b) Coordinate systems of Shmoobot.
• Figure 3: (a) Definition of contact frame. (b) Schematic of contact.
• Figure 4: Control framework diagram. A reference trajectory is first generated by a path planner or sent by the user. A bi-level MPC will calculate an optimal trajectory that tracks the reference. The environment SDF used by the controller needs to be recomputed. The upper level of the controller (the blue block) is a contact-implicit MPC which generates a draft of the motion plan with soft contact models. The lower level of the controller (the yellow block) is a hybrid MPC. It will extract a contact schedule from the motion plan, and refine the trajectory with hard contact models. The low level balancing controller and arm controller will then track the motion plan provided by the hybrid MPC.
• Figure 5: (a) Key frames for the disturbance rejection experiment. The end effectors are highlighted with blue and red circles. A red circle indicates that the end effector is in contact. At T = 0.5 s, a human operator pushed the robot to the wall. The robot actively used its arm to recover from the disturbance. (b) Force output trajectory, acceleration trajectory and mode sequence of the robot.