Grégoire Passault

Metabot: A Low-Cost Legged Robotics Platform for Education

This paper introduces an open-source 3D printed low-cost legged robot environment designed for educational purposes. The platform, called Metabot, is already operational and is currently experienced by teachers and pupils. The robot as a teaching aid, has several advantages: at first, it is stimulating for pupils, especially legged robots. Second, robotics is multidisciplinary and centralizes a broad spectrum of knowledge for the pupils, in particular, it allows to introduce programming in a concrete environment, which is now an important need in school. Finally, legged robots can illustrate concretely several matters of geometry.

Low-cost force sensors for small size humanoid robot

Summary form only given. We present a new design of foot pressure sensors in the context of small low-cost robots. We design the robot in the context of RoboCup kid-size humanoid league (i.e. the robot size must be lower than 90 cm). A new challenge is to walk on artificial grass of 3 cm height, which essentially means a soft irregular floor. In this context, one cannot use full 6-axis force sensors for cost and mechanical integration reasons. Classically, FSR sensors are used to handle foot pressure with unclear efficiency. Instead of that, we propose to use low cost strain gauge originally designed for scales. The strain gauge is made of resistors network glued on a aluminium beam. In our design, the aluminium beams are themselves full parts of the mechanical structure of the foot. Therefore, we measure directly the deformation of the foot and thus the force applied on it in a robust way. We use this system in order to measure the position of the centre of pressure under the foot, in order to balance motor primitives of the robot: perturbation rejection, locomotion, shoot. First experiments are promising and we plan to set up a rigorous comparison with the ZMP.

Dynaban, an Open-Source Alternative Firmware for Dynamixel Servo-Motors

In this paper, we present an alternative open-source firmware for the Dynamixel MX-64 servo-motor. We discuss software features to fully exploit the hardware capabilities of the device. In order to enhance the default controller, a friction model and an electric model of the motor are embedded into the firmware. The parameters of the model are found using a black-box optimization algorithm. A feed-forward method is proposed to follow position, speed and torque trajectories. The approach is tested with a highly dynamic kick movement on our humanoid soccer robot Sigmaban whose torque trajectories are computed using a classic rigid body inverse dynamics. The comparison between the default control strategy and the proposed one shows significant improvements in terms of accuracy, delay and repeatability.

Learning the odometry on a small humanoid robot

Odometry is an important element for the localization of mobile robots. For humanoid robots, it is very prone to integration errors, due to mechanical complexity, uncertainties and foot/ground contacts. Most of the time, a visual odometry is then used to encompass these problems. In this work we propose a method to compensate for odometry drifting using machine learning on a small size low-cost humanoid without vision. This method is tested on different ground conditions and exhibits a significant improvement in odometry accuracy.

Optimizing Morphology and Locomotion on a Corpus of Parametric Legged Robots

In this paper, we describe an optimization approach to the legged locomotion problem. We designed a software environment to manipulate parametrized robot models. This environment is a platform developed for future experiments and for educational robotics purpose. It allows to generate dynamic models and simulate them using a physics engine. Experiments can then be made with both morphological and controller optimization. Here we describe the environment, propose a simple open loop generic controller for legged robots and discuss experiments that were made on a robot corpus using a black-box optimization.

A Fully Autonomous Robot for Putting Posts for Trellising Vineyard with Centimetric Accuracy

We present an autonomous robot designed to optimize and to make secure the process of putting vineyard post during trellising. The principle is as follows: The robot goes at the location of the post in an autonomous way with a centimetric accuracy, the user provides the post, and then the robot drives the post into earth without additional user intervention. The robot is designed from a small backhoe loader. The tool of the backhoe loader has been replaced by a hydraulic hammer. All the hydraulic functions of the machine are controlled, including the hammer and the tracks. The robot is geolocalized with a differential global navigation satellite systems (GNSS for short) mono-frequency low-cost system relying on a mix of triple and double differences.

An experiment of low cost entertainment robotics

This paper reports about the robotic installation set up by the Rhoban Project in the French pavilion of the Expo 2012 of Yeosu, Korea ([6]). The installation has consisted in a humorous show involving humanoid robots and anthropomorphic arms, with the illusion of life as a guideline. We emphasized natural compliant motion and physical interaction in order to make the show attractive. The design raised some issues dealing with robustness of robots, but also with the realism of the motions and the synchronization of the robots with the music.

Rhoban Football Club: RoboCup Humanoid Kid-Size 2017 Champion Team Paper

In 2017, Rhoban Football Club reached the first place of the Kid-Size soccer competition for the second time, and was awarded best humanoid by the jury (https://www.robocuphumanoid.org/hl-2017/results/.). Capitalizing over the hardware design and software basis from previous participations, we improved both robustness and overall behaviors of our humanoid robots. An interesting milestone reached is the opportunistic pass ability; since such an high-level team play behavior can only work after solving underlying difficulties like having a good field localization, robust and reliable detection of vision features, a repeatable locomotion with odometry etc. This paper describes the most meaningful improvements accomplished over this year along with perspectives of future work.