Special gimbal mechanisms and algorithms that implement decentralized compliant control have been developed for use in research on the sensors, the actuators, and the design and functional requirements for systems of multiple mobile robots cooperating in site-clearance and construction operations. The gimbal mechanisms and control algorithms were designed, in particular, to enable two robotic exploratory vehicles (i.e., rovers) to transport a long payload while moving along the ground in a commanded formation. Although these developments are parts of a continuing effort to develop robotic capabilities for exploration of Mars, the same robotic capabilities could be expected to find application on Earth.

Figure 1. The Gimbal Mechanism enables the gripper to move freely in one translational and three rotational degrees of freedom, and measures forces and torques.

Each

Figure 2. Two Mars Rovers transport a long payload in diagonal formation.

gimbal mechanism (see Figure 1) has four degrees of freedom. One such mechanism is part of each rover. The gimbal incorporates a compliant gripper on a longitudinal slider for “soft grip” of a payload. The gimbal is passive and is fully instrumented with potentiometers to measure the orientation and position (pitch, roll, yaw, and lateral translation) of the gripper. The gimbal mechanism is mounted on a six-degree-of- freedom load cell, which is used to resolve reaction forces. The load cell, in turn, is mounted on a cross brace between shoulders of the robotic vehicle.

The decentralized compliant control scheme uses no explicit communications; i.e., the rovers do not “talk” to each other via wireless modems but communicate with each other implicitly via their common payload through force sensors. The scheme involves four low- level behaviors denoted formation controller, minimize forces/torques on payload, center payload in longitudinal slider, and group formation. The control inputs for three of the behaviors are the speed and heading of a rover. The formation controller behavior receives a formation-angle command from the group formation behavior. The commanded formation angle is mapped to the corresponding gimbal yaw angles on the two rovers. The formation controller behavior then seeks to control the speed and heading of each rover in an effort to achieve and maintain the commanded gimbal yaw angle on each rover.

The minimize forces/torques on payload behavior seeks to minimize the forces on the payload or compliant linkage on each rover. The forces on the payload can be high if the relative speed between the two rovers is greater than a set threshold. The magnitude of the force along the longitudinal axis of the payload is the input for this behavior. The predominant control output of the minimize forces/torques on payload behavior is a rover speed command, supplemented with steering-correction commands.

The center payload in longitudinal slider behavior seeks to minimize deviations of the payload from midpoint of the longitudinal slider on each rover. The control outputs of the center payload in longitudinal slider behavior are a rover-speed and heading (steering) control command.

Proportional-plus-derivative (PD) controllers for speed and heading modifications that satisfy the requirements for the formation controller and center payload in longitudinal slider behaviors under steady-state conditions have been developed. The PD controllers independently achieve their respective goals, but when implemented simultaneously, they give conflicting speed and heading corrections. To resolve these conflicts, the outputs of the PD controllers are combined by use of a weighting scheme to compute speed and heading corrections for each rover.

In several experiments performed at Arroyo Seco in Pasadena, California, the following actions were demonstrated:

  • A pair of Mars rovers compliantly coupled to a common payload (see Figure 2) autonomously moved, variously, forward or backward through distances of 5 to 50 m over uneven, natural terrain.
  • The pair of rovers compliantly coupled to a common payload autonomously changed formations between arbitrary initial and final formations (including row, column, and diagonal formations).

This work was done by Ashitey Trebi-Ollennu, Hari Das, Anthony Ganino, Hrand Aghazarian, and Brett Kennedy of Caltech for NASA’s Jet Propulsion Laboratory.