Content

Air-Ground Cooperation without Global Information: RoFly and CubeTrack Cooperation with CREPES and CoNi-MPC

Authors:

Jiadong Lu, Jiajun Yu, Li Wang, Mingxuan Zhang, Pengxiang Zhou, Ruitian Pang, Xiangyu Li, Zhehan Li, Chao Xu, Yanjun Cao

• Segment 1: Rofly
• RoFly is a novel passive-wheeled Transformable Aerial- Bipedal Vehicle (TABV) that features a pitch-torque balanced transformable parallelogram linkage mechanism combined with bidirectional rotor control, achieving high energy ef- ficiency in terrestrial mode. The work includes a bimodal dynamic model of the proposed TABV with corresponding differential flatness to simplify trajectory planning. A Hybrid Nonlinear Model Predictive Control (HNMPC) controller is developed to achieve accurate trajectory tracking in both terrestrial and aerial modes, enabling seamless transitions between them. Extensive experimental results validate the energy efficiency and control performance of the proposed system

• Segment 2: CubeTrack
• CubeTrack is an innovative reconfigurable tracked robot whose core innovation is the Geometry-Changing Track Module (GCTM) featuring our novel Quad-slider elliptical trammel mechanism (Qs-ETM). This mechanism enables dy- namic posture adjustment of the flipper arms along optimized elliptical trajectories to overcome various obstacles: when climbing vertical barriers up to 0.33m high, the flippers rotate downward to form a stable triangular configuration; when crossing gaps up to 0.6m wide, they extend bidirectionally to bridge the chassis; and during stair descent, they ensure smooth weight transition. The modular architecture integrates a central chassis with two GCTMs containing direct-drive motors for precise control, allowing effective operation in both compact triangular and flat configurations according to terrain demands.
  Supported by our self-developed trajectory planning al- gorithms and kinematic modeling framework based on in- stantaneous centers of rotation, CubeTrack achieves fully autonomous navigation in complex multi-layer architectural environments. Our proprietary motion planning system en- ables the robot to generate optimal trajectories in real-time while accounting for nonholonomic constraints and terrain variations. Experimental results demonstrate that our method empowers CubeTrack to successfully navigate challenging non-Manhattan environments featuring spiral staircases, nar- row passages, and multi-level structures, demonstrating sig- nificant advantages of our innovative approach.

• Segment 3: CREPES
• For an efficient multi-robot system, mutual relative lo- calization is the key to accomplishing tasks cooperatively. Stable, accurate and fast relative pose estimation between robots can significantly improve the quality of collaboration. CREPES is a collaborative relative pose estimation system for multi-robot systems, which can provide accurate and real- time 6-DoF relative pose estimation between robots. The system is designed requiring no external equipment or prior knowledge of the environment. Each robot is equipped with a low-cost MEMS IMU, a UWB module, a camera with a fisheye lens and IR LEDs for mutual identification and bearing measurement. The system fuses the measurements from these sensors through an Error State Kalman Filter (ESKF) to estimate the relative poses between robots. The system can run at a frequency of 100 Hz with centimeter- level translation accuracy and degree-level rotation accuracy in real-time.

• Segment 4: Coni-MPC
• CoNi-MPC presents a novel solution for UAV control in cooperative multi-robot systems, which can be used in various scenarios such as leader-following, landing on a moving base, or specific relative motion with a target. Unlike classical methods that tackle UAV control in the world frame, we directly control the UAV in the target coordinate frame, without making motion assumptions about the target. In detail, we formulate a non-linear model predictive controller of a UAV, referred to as the agent, within a non-inertial frame (i.e., the target frame). The system requires the relative states (pose and velocity), the angular velocity and the accelerations of the target, which can be obtained by relative localization methods and ubiquitous MEMS IMU sensors, respectively. This framework eliminates dependencies that are vital in classical solutions, such as accurate state estimation for both the agent and target, prior knowledge of the target motion model, and continuous trajectory re-planning for some complex tasks.