Toward Understanding Key Estimation in Learning Robust Humanoid Locomotion

Accurate state estimation plays a critical role in ensuring the robust control of humanoid robots, particularly in the context of learning-based control policies for legged robots. However, there is a notable gap in analytical research concerning estimations. Therefore, we endeavor to further understand how various types of estimations influence the decision-making processes of policies. In this paper, we provide quantitative insight into the effectiveness of learned state estimations, employing saliency analysis to identify key estimation variables and optimize their combination for humanoid locomotion tasks. Evaluations assessing tracking precision and robustness are conducted on comparative groups of policies with varying estimation combinations in both simulated and real-world environments. Results validated that the proposed policy is capable of crossing the sim-to-real gap and demonstrating superior performance relative to alternative policy configurations.

RING++: Roto-translation Invariant Gram for Global Localization on a Sparse Scan Map

Global localization plays a critical role in many robot applications. LiDAR-based global localization draws the community's focus with its robustness against illumination and seasonal changes. To further improve the localization under large viewpoint differences, we propose RING++ which has roto-translation invariant representation for place recognition, and global convergence for both rotation and translation estimation. With the theoretical guarantee, RING++ is able to address the large viewpoint difference using a lightweight map with sparse scans. In addition, we derive sufficient conditions of feature extractors for the representation preserving the roto-translation invariance, making RING++ a framework applicable to generic multi-channel features. To the best of our knowledge, this is the first learning-free framework to address all subtasks of global localization in the sparse scan map. Validations on real-world datasets show that our approach demonstrates better performance than state-of-the-art learning-free methods, and competitive performance with learning-based methods. Finally, we integrate RING++ into a multi-robot/session SLAM system, performing its effectiveness in collaborative applications.

LiDAR-Inertial 3D SLAM with Plane Constraint for Multi-story Building

The ubiquitous planes and structural consistency are the most apparent features of indoor multi-story Buildings compared with outdoor environments. In this paper, we propose a tightly coupled LiDAR-Inertial 3D SLAM framework with plane features for the multi-story building. The framework we proposed is mainly composed of three parts: tightly coupled LiDAR-Inertial odometry, extraction of representative planes of the structure, and factor graph optimization. By building a local map and inertial measurement unit (IMU) pre-integration, we get LiDAR scan-to-local-map matching and IMU measurements, respectively. Minimize the joint cost function to obtain the LiDAR-Inertial odometry information. Once a new keyframe is added to the graph, all the planes of this keyframe that can represent structural features are extracted to find the constraint between different poses and stories. A keyframe-based factor graph is conducted with the constraint of planes, and LiDAR-Inertial odometry for keyframe poses refinement. The experimental results show that our algorithm has outstanding performance in accuracy compared with the state-of-the-art algorithms.

Multi-expert learning of adaptive legged locomotion

Achieving versatile robot locomotion requires motor skills which can adapt to previously unseen situations. We propose a Multi-Expert Learning Architecture (MELA) that learns to generate adaptive skills from a group of representative expert skills. During training, MELA is first initialised by a distinct set of pre-trained experts, each in a separate deep neural network (DNN). Then by learning the combination of these DNNs using a Gating Neural Network (GNN), MELA can acquire more specialised experts and transitional skills across various locomotion modes. During runtime, MELA constantly blends multiple DNNs and dynamically synthesises a new DNN to produce adaptive behaviours in response to changing situations. This approach leverages the advantages of trained expert skills and the fast online synthesis of adaptive policies to generate responsive motor skills during the changing tasks. Using a unified MELA framework, we demonstrated successful multi-skill locomotion on a real quadruped robot that performed coherent trotting, steering, and fall recovery autonomously, and showed the merit of multi-expert learning generating behaviours which can adapt to unseen scenarios.

Search-based Kinodynamic Motion Planning for Omnidirectional Quadruped Robots

Autonomous navigation has played an increasingly significant role in quadruped robots system. However, existing works on path planning used traditional search-based or sample-based methods which did not consider the kinodynamic characteristics of quadruped robots. And paths generated by these methods contain kinodynamically infeasible parts, which are difficult to track. In the present work, we introduced a complete navigation system considering the omnidirectional abilities of quadruped robots. First, we use kinodynamic path finding method to obtain smooth, dynamically feasible, time-optimal initial paths and added collision cost as a soft constraint to ensure safety. Then the trajectory is refined by timed elastic band (TEB) method based on the omnidirectional model of quadruped robot. The superior performance of our work is demonstrated through simulated comparisons and by using our quadruped robot Jueying Mini in our experiments.

Bound Controller for a Quadruped Robot using Pre-Fitting Deep Reinforcement Learning

The bound gait is an important gait in quadruped robot locomotion. It can be used to cross obstacles and often serves as transition mode between trot and gallop. However, because of the complexity of the models, the bound gait built by the conventional control method is often unnatural and slow to compute. In the present work, we introduce a method to achieve the bound gait based on model-free pre-fit deep reinforcement learning (PF-DRL). We first constructed a net with the same structure as an actor net in the PPO2 and pre-fit it using the data collected from a robot using conventional model-based controller. Next, the trained weights are transferred into the PPO2 and be optimized further. Moreover, target on the symmetrical and periodic characteristic during bounding, we designed a reward function based on contact points. We also used feature engineering to improve the input features of the DRL model and improve performance on flat ground. Finally, we trained the bound controller in simulation and successfully deployed it on the Jueying Mini robot. It performs better than the conventional method with higher computational efficiency and more stable center-of-mass height in our experiments.