Search and rescue operations are extremely dangerous, where even trained men and women face a high risk of injury or death in their effort to save lives. One solution in the face of disasters caused by fires, earthquakes and the like, is designing robots that can reach areas too dangerous for humans. While there are a number of different potential designs, one of the most interesting is a four legged robot being designed by the Dynamic Legged Systems Lab led by Dr. Claudio Semini (Italian Institute of Technology) to perform dynamic tasks such as walking in rugged terrain, carrying heavy loads, climbing hills and even jumping and running. Such rescue robots can supplement scarce human searchers in order to locate and rescue victims faster, a valuable resource in situations where speed can help save lives.
One of the critical design challenges faced by IIT engineers is the selection of the bearings and beams used in the legs. The stiffness of the bearings and beams governs the value of the forces on the hip and knee. The link flexibility problem can be solved analytically with equations, but this approach is time consuming and requires a high level of mathematical skill. The complexity of the analytical method increases exponentially as the degrees of freedom and geometric complexity of the robot increase.
Instead, Advanced Industrial Automation Lab engineers used Adams and MSC Nastran to simulate the performance of the robot in a fraction of the time. Each robot leg contains about 450 parts. IIT engineers merged these components into just a few rigid bodies, one for each moving part of the structure. These included the slider, upper leg, lower leg, hip cylinder, hip beam, knee cylinder and knee beam as shown in the diagram above. The position of each rigid body was characterized by three reference points, two on the ends of each body and one in the center of mass. The actuators’ movement was modeled with spring-damping elements.
Two different models were built from these elements. The first, called the Rigid Connected Prototype (RCP), uses rigid connections with infinite stiffness at the joints between the elements. The second, called the Flexible Connected Prototype (FCP), uses flexible joints created with Adams bushing elements.
The accuracy of the two models was evaluated with a drop test. At the beginning of the test the leg was standing in its equilibrium position. A vertical guide was used to maintain vertical alignment during the test. Then the leg was raised to a fixed elevation, dropped and left free to oscillate from the impact. During the test the slider position was measured with an absolute encoder and the forces on the knee and hip joints were measured with load cells. The experiment was repeated three times so that its repeatability could be gauged.
The experimental results were compared to Adams simulation results. In the RCP simulation, the slider displacement shows good correlation with the experimental results. However, the hip and knee forces predicted by simulation were considerably higher than those seen in the physical experiment. This is explained by the unrealistically high stiffness in the joints of the RCP model. Two variations of the FCP simulation were performed. One variation used four Adams bushing elements corresponding to joints 1, 2, 4 and 6 in the diagram. The second variation used six Adams bushing elements at all six joints in the diagram. The FCP simulation provides results that correlate well with the physical test results for both forces and displacement. The FCP simulation with four bushings shows excellent correlation engineers will begin evaluating a wide range of design alternatives in order to improve the performance of the robot. When the design has been optimized in the simulation world, then a new prototype will be constructed to validate these improvements in the real world.