CERBERUS: The AI Based Quadruped Robot
CERBERUS is a quadruped robot with 3 joints per leg (a total of 12) which makes it flexible for rough terrain. Giving remote operations a boost, this robot can be controled by a web application through which the robot receives the commands and uses Reinforcement Learning to generate trajectory for i
2025-06-28 16:30:46 - Adil Khan
CERBERUS: The AI Based Quadruped Robot
Project Area of Specialization RoboticsProject SummaryCERBERUS is a quadruped robot with 3 joints per leg (a total of 12) which makes it flexible for rough terrain. Giving remote operations a boost, this robot can be controled by a web application through which the robot receives the commands and uses Reinforcement Learning to generate trajectory for its 12 joints. It can also balance itself if given a slight push. It uses minimal computational hardware which makes it cheaper from its competitors and also scalable for further attachments.
The objective is to increase the use of quadruped robots in modern industry due to its flexible mobility by making it more intelligent. As pure mathematical control models for locomotion of quadruped robots are complex and computationally expensive, our goal is to use Reinforcement Learning method to achieve an alternate and computationally cheaper model to control the locomotion of our robot. The objective also extends as our goal is also to reduce the time required and hardware damage by training the Reinforcement Learning model through simulation environment using physics engine before deploying it on the actual physical robot.
Project Implementation MethodThe sequence of steps for implementation is:
1. Making a 3D model and URDF file
2. 3D printing and assembly of the model with selected electronics
3. Importing the URDF to Bullet Physics Engine
4.Training the PPO (Reinforcement Learning Algorithm by OpenAI) using the imported URDF
5. The learned policy (Reinforcement Learning) will then be deployed on Jetson Microcontroller
6. A web portal will be designed through which the robot will be controlled
7. The robot will use commands from web portal to perform tasks through the trained Algorithm and a gyro to balance itself
Benefits of the ProjectAs our objective is to increase the use of quadruped robots in modern industry there are many advantages that comes with it some of them are:
1. It can be used as a personal companion/assistant in home and workplaces and perform basic tasks.
2. A swarm of these robots can be used to help perform operations in mines and other risky places due to their mobility and control options.
3. They can be used for security surveillance and patrolling and give live feed/alerts on centralized systems.
Technical Details of Final Deliverable1. Locomotion of robot through reinforcement, trained on simulation environment.
2. Locomotion and balancing of the real robot.
3. Control of the robot through a web application.
4. Following modes of control of the robot:
i. Stand
ii. Walk
iii.Run
iv. Stop
Final Deliverable of the Project HW/SW integrated systemCore Industry ITOther Industries Manufacturing , Security Core Technology Artificial Intelligence(AI)Other Technologies RoboticsSustainable Development Goals Industry, Innovation and InfrastructureRequired Resources| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 69850 | |||
| NVIDIA Jetson Nano Developer Kit | Equipment | 1 | 22000 | 22000 |
| MPU6050 Sensor | Equipment | 2 | 200 | 400 |
| 3D Printed Model | Equipment | 1 | 25000 | 25000 |
| MG996R Servo | Equipment | 12 | 750 | 9000 |
| Damaged Hardware Replacement | Miscellaneous | 1 | 5000 | 5000 |
| PCA9685 Servo Motor Driver | Equipment | 1 | 450 | 450 |
| 2x 11 V 3 Cell Lipo Battery 4000 mAh | Equipment | 2 | 4000 | 8000 |