Abstract
We present a novel outdoor navigation algorithm to generate stable and efficient actions to navigate a robot to the goal. We use a multi-stage training pipeline and show that our model produce policies that result in stable and reliable navigation of robots on complex terrains. Based on the Proximal Policy Optimization (PPO) algorithm, we developed a novel method to achieve multiple capabilities for outdoor navigation tasks, namely: alleviate robot drifting, keep robots stable on bumpy terrain, avoid climbing on hills with steep elevation changes, and avoid obstacles in navigation. Our training process mitigates the reality(sim-to-real) gap by introducing more generalized environmental and robotic parameters and training with rich features of Lidar perception in the Unity simulator. We evaluate our method in both simulation and the real world on Clearpath Husky and Jackal. Additionally, we compare our method against the state-of-the-art approaches and show that in the real world it improves stability on hilly terrain by at least $30.7\%$, reduces drifting by $8.08\%$, and for high hills our trained policy can prevent the robot move on area with high gradient and further keep small change of the elevation of the robot in each step of motion.
Abstract (translated)
URL
https://arxiv.org/abs/2205.03517