Impact Aware Aerial Robotics

These days robots are increasingly interacting with their environment. However, there is still a lot of room for improvement in improving the efficiency of interaction tasks. Most applications perform contact tasks at near-zero velocities in order to be able to preserve the robot's stability. The dynamic contact transition problem studies contact tasks that take advantage of impacts and establish contacts at relatively higher velocities. It can increase task execution efficiency, reduce net energy consumption and increase the repertoire of tasks a robot is capable of performing. This problem has been effectively applied to manipulators and humanoids. This project explores the possibility of extending dynamic contact transition to aerial robots.

Aerial robots are a growing presence in the robotics industry owing to their unique vantage offering access to remote and dangerous environments. Most aerial robotics systems deployed today suffer from short flight times owing to limited power capabilities. Additionally, interaction with the environment is complicated by the problems encountered when performing tasks from a floating base. Overcoming these difficulties usually results in compliant control algorithms for lower impact velocities or complicated hardware solutions that increase the cost incurred and energy used. Successfully implementing dynamic contact transition in aerial robots offers a possibility to address some of these issues by enabling simpler systems like quadrotors to perform interaction tasks like swooping, intercepting a target in mid-air or force application.

The dynamic contact problem is explored here through the use case of swooping motion, widely seen in birds. To develop a test bed simulation scenario, key features of the motion are identified and translated into a trajectory-tracking task for the quadrotor. Three different scenarios of impact events are modelled based on the design of the object being impacted and the task executed.

The proposed controller extends the range of desired velocities where the quadrotor performs the trajectory tracking task utilising the dynamic contact transition problem. It is inspired by the Interaction planning problem and makes use of robotic reflex reactions. The scope of the proposed solution is set to high-impact tasks at low altitudes on systems with input saturation. The hypothesis that altitude factors into the task's success is confirmed through some high-altitude simulation tests. Furthermore, an error space is introduced that distinguishes post-impact behaviour where the quadrotor successfully executes the dynamic contact transition task and instances where it fails. The proposed controller is observed to increase the region in error space where the quadrotor succeeds. Finally, exploring the impact-variant subspace of the problem highlights some challenges involved in adapting dynamic contact transition control techniques to underactuated aerial robots.

The contributions discussed in this report satisfy the research goals laid out at the beginning of this project. The dynamic contact transition problem is explored from the lens of a specific use case on aerial robots. The exploration was limited to simulations, an ideal next step would be to validate the proposed solution through physical experiments.