SigmaBan has been designed after Acroban in order to optimize the ratio power/weight and still explore morphology. It is a lightweight robot, about 50cm tall for 4 kg. It is fully autonomous, in particular it embeds vision and energy. We have payed a particular attention to the robustness in such a way that it can fall and get up without breaking. This implied numerous difficult engineering issues. However beyond that, this allows to develop and adjust motor primitives by direct experiment on the robot itself. Therefore the differences between the modeling and the concrete platform is no more a problem.

SigmaBan test bench

We use SigmanBan to conduct numerous experiments on walking. For instance, by using a motion capture system, we tried to enforce learning methods in order to learn odometry of the robot.

A very early version of SigmaBan walking and getting up

Even if there has been several versions of the robot, originally it has been used to investigate structural compliance in mechanics: the use of passive joints with dampers at different levels of the structure (spine and hip). Even if we like motors, passive walkers are a continuous inspiration source in our designs.

Experiment on mechanical structure using dampers
A long term work drove us to get a complete modular control architecture, the rhoban control system. It allows to develop robot behaviors at all levels of control, from motor primitives to high level image processing. It provides a graphical environment giving a global view of motor primitives, but it also allows to get motor primitives by direct physical demonstrations on the robot itself.

Direct motion record

Rhoban Control Environment