Manual, Semi or Fully Autonomous Flipper Control? A Framework for Fair Comparison

TL;DR

The paper benchmarks manual, semi-autonomous, and autonomous flipper control for skid-steer robots by reimplementing baselines on a common platform and introducing a domain-aware semi-autonomous policy. It defines novel cognitive load and traversal quality metrics and provides a benchmarking interface to generate Quality-Load graphs. Results in a 2D Quality-Load space show the proposed policy bridges the gap between autonomous and manual control, achieving high traversal quality with lower cognitive load. Surprisingly, an experienced operator using full manual control can outperform autonomous methods on complex terrains, while the best autonomous approach can closely approach manual quality with significantly reduced workload. Overall, the work offers a practical framework for fair comparison and advances in domain-informed hybrid control for robust terrain traversal.

Abstract

We investigated the performance of existing semi- and fully autonomous methods for controlling flipper-based skid-steer robots. Our study involves reimplementation of these methods for fair comparison and it introduces a novel semi-autonomous control policy that provides a compelling trade-off among current state-of-the-art approaches. We also propose new metrics for assessing cognitive load and traversal quality and offer a benchmarking interface for generating Quality-Load graphs from recorded data. Our results, presented in a 2D Quality-Load space, demonstrate that the new control policy effectively bridges the gap between autonomous and manual control methods. Additionally, we reveal a surprising fact that fully manual, continuous control of all six degrees of freedom remains highly effective when performed by an experienced operator on a well-designed analog controller from third person view.

Figures (5)

• Figure 1: Cognitive load vs traversal quality: State-of-the-art methods outlined by grey clusters, the proposed method outlined by the red cluster. All methods manually control the speed and heading of the robot, which counts against the cognitive load. $\bullet$ Semi-autonomous flipper control [Ours], $\bullet$ Manual flipper control - discrete modes Larochelle-IJRA-2013, modes Larochelle-IJRA-2013 + anti-stuck Teymur2022, $\bullet$ Manual flipper control - continuous, $\bullet$ Autonomous flipper control - discrete modes Larochelle-IJRA-2013 + anti-stuck Teymur2022, $\bullet$ Autonomous flipper control - continuous pan-2023.
• Figure 2: Robot used for experimental evaluation: Left shows our custom-built skid-steer robot with four independently controlled flippers. The right column displays our simulator, which employs a belt simulation instead of the multiple wheels commonly used in publicly available simulators.
• Figure 3: Xbox Series X Controller mapping: L-Stick controls ride movement, while R-Stick controls flipper tilting when combined with L1 (front left), R1 (front right), L2 (rear left), or R2 (rear right) buttons.
• Figure 4: Typical traversal modes: The figure is presented with the permission of authors Teymur2022.
• Figure 5: Gazebo Testing arena: arena implemented in simulator Gazebo Gazebo consisting of 13 various obstacles including an up-down staircase Teymur2022.