Vision based Smart Decisions for MultiTerrain WhegRoach Robot

This project proposes a cockroach-like bio-inspired robot, called WhegRoach robot. Its name WhegRoach is derived from Wheg and cockRoach. By Wheg it means, it has C-shaped legs, the combination of wheels and legs, while its gaits and body design is inspired from cockroach because it can stabilize it

Project Title

Vision based Smart Decisions for MultiTerrain WhegRoach Robot

Project Area of Specialization

Robotics

Project Summary

This project proposes a cockroach-like bio-inspired robot, called WhegRoach robot. Its name WhegRoach is derived from Wheg and cockRoach. By Wheg it means, it has C-shaped legs, the combination of wheels and legs, while its gaits and body design is inspired from cockroach because it can stabilize itself at any rough surface while moving at high speed. Its design allows it to move on uneven, flat surfaces and terrains even under challenging environmental conditions. The passive compliance in legs overcome the limitations of under actuation and helps to simplify the mechanical design and increase robustness.

A WhegRoach may have different numbers of legs ranging from four to thirty. More Whegs provide more payload capacity and unique gaits can be generated but at the same time it requires more power and control becomes complex and vice versa. We will be focusing on a 6-Wheg robot, as it has enough Whegs to generate multiple gaits, provide appropriate payload capacity and has less complexity (only 6 DoF). Each Wheg is controlled by a separate DC Motor, which allows it to generate different gaits easily and increase its power and stability. Its design allows it to move upside down providing it a great flexibility.

For integrating intelligence and for making it to perform actions autonomously, a vision sensor (Kinect Vision Sensor) will be used. With the help of a vision sensor, robot will sense and perceive the outside environment and take decisions accordingly. If an obstacle happens in its way, it will change its course or if the robot observes a stair, it will change its gait accordingly to climb it. Furthermore, if it observes a gap, the robot will try to jump over it. The vision algorithms help the robot to decide which predefined gait it should select for a particular scenario.

Project Objectives

This project focuses on the hardware and software implementation of the WhegRoach robot, leading from CAD modelling to the final Hardware product. Main objectives of this project are detailed as follow:

1. 3D modelling and simulation of the robot in CoppeliaSim (a physics-engine for robotics simulations).
2. Bridging the CoppeliaSim framework to the remote API’s for the control of the robot.
3. Implementation of Tripod and Stair-Climb locomotive gaits and testing in CoppeliaSim.
4. Integration of Vision and other sensors with the robot in the simulation framework.
5. Implementation of Vision-based smart decisions such as stair climbing, gap overcome and obstacle avoidance using computer-vision methods in the simulation framework.
6. Development of a test-bed to test the robot performance in various environmental conditions in the simulation framework.
7. Implementation and testing of all the above-mentioned objectives on actual hardware.

Project Implementation Method

3D CAD Modeling:

For CAD Modeling of our robot, we will be using SolidWorks. The purpose of using it is because, it is very flexible, handy software and provides a variety of tools to test the mechanical design of the robot. Furthermore, we can easily import ‘.stl’ files in much robotic software. Leg and body design, its dimensions will be set in here and then will import this file to the robotic simulator.

Robotics Simulator:

CoppeliaSim is a physics-engine for robotics simulations. It has a free educational version, providing almost all tools that are required for real-time simulation. We can make pure and non-pure objects using infinite triangles and set parameters like mass, the density of the robot to closely match the real-world robot. We will import the CAD design of our robot in it and then make its pure, non-pure objects and joints. Generation of gaits, making mock environment and controlling of vision sensor will be done in CoppeliaSim with the help of coding. For coding, Python language will be used to make logics(detect an obstacle, its height, distance) using external API’s that CoppeliaSim provides. OpenCV library will be used for vision algorithms.

Hardware Requirements:

Last, we will switch from simulation to real-world implementation and in the hard form, we will repeat the same above steps. We will 3D print the SolidWorks CAD model using a 3D printer. For controlling and processing, we will be using Jetson Nano, DC Motors, and Kinect Vision Sensor. Programming will be done using Python3.

The following figure is showing the methodology(workflow) of our project.                                    

                                           Figure 1: Methodology 

Benefits of the Project

This project focuses on an autonomous robot capable of making smart decisions and giving an internal view to the rescue team so that they can go safely and perform their respective tasks. The design of the robot enables it to overcome different obstacles and move stably on rough terrains. It can be used for space exploration, crops field, Search and Rescue(SAR) tasks and mining areas to perform perilous tasks. For military, space companies, search and rescue teams, it would be of great interest.

Technical Details of Final Deliverable

The WhegRoach comprising all of its gaits and vision integration will be able to take self-decisions and reduce human interaction and effort. With the help of Kinect sensor, the robot will take the depth images and by applying colour coding, masks, image segmentation technique, the robot will estimate the distance of the obstacle and change its path. Similarly, for the gap, it will observe whether the gap-distance is in its range or not and if so, it will select the predefined gait to overcome it. For stairs, the robot will then again choose the stair-climbing gait and will climb it with ease and stability. The depth images from Kinect sensor will be received by Jetson Nano and with the help of OpenCV library, the controller will process the images and result will be sent to the robot joints and it will then act accordingly. All this process will be in real-time. The robot will be simulated first in CoppeliaSim and then It will be interface with the hardware with the help of ROS (Robot Operating System). The CAD model will be designed in SolidWorks cause of its great compatibility and ease of use. In Hardware, the body, and legs, designed in SolidWorks, will be printed using the 3D printer. Then the motors, Kinect sensor and Jetson Nano will be attached. Then there comes the coding, controlling, and interfacing of the robot with the simulator. The end product when presented will be intelligent enough to take self-decisions and perform its respective tasks and have plenty of future work also like with the help of Reinforcement Learning(RL) the robot will learn path planning and reach its goal.

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Others

Other Industries

Agriculture , Security

Core Technology

Others

Other Technologies

3D/4D Printing, Robotics

Sustainable Development Goals

Decent Work and Economic Growth

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com