Group 5
Police Chasing Robot
Objective:
Develop a robot that can identify and chase a specific car based on a description, using deep learning techniques for object detection and tracking, and implement appropriate actions through robotics.
Deep Learning:
Car Identification:
Description-Based Detection: The system should identify a specific car based on given descriptions such as color (e.g., "a blue car") or make (e.g., "a Honda").
Model Choice: Utilize YOLO (You Only Look Once) for real-time object detection.
Bounding Box Differentiation: Highlight the identified car with a different colored bounding box for easy tracking.
Trajectory Prediction:
Efficiency Improvement: Implement trajectory prediction using RNN (Recurrent Neural Network) or transformer models to enhance the efficiency and accuracy of YOLO.
Model Integration: Integrate the trajectory prediction model with YOLO to anticipate the car's movement and adjust the tracking accordingly.
Implementation Steps:
Dataset Collection: Gather a dataset of car videos with various descriptions for training the YOLO model.
Model Training: Train YOLO for car detection and integrate RNN or transformer models for trajectory prediction.
Testing: Validate the system's accuracy and real-time performance with test videos.
Robotics:
Action Proposal:
Possible Actions: Determine the appropriate actions the robot should take upon identifying the target car, such as following the car or shooting (depending on the intended non-lethal design).
Actuator Design: Design actuators that can perform the proposed actions. Ensure the actuators are functional and safe.
Camera Mobility:
Rotating and Tracking: Equip the robot with a rotating camera capable of following the identified car.
Autonomous Navigation: Ensure the robot can move around autonomously while keeping the target car in view.
Implementation Steps:
Camera System: Develop a camera system that can rotate and follow the target car accurately.
Actuation Mechanism: Design and test actuators for the proposed actions, ensuring they work effectively with the camera system.
Integration: Ensure the camera and actuator systems work seamlessly with the deep learning model for real-time tracking and action execution.
Next Steps for Students:
Deep Learning Tasks:
Data Collection: Collect videos of cars with various descriptions.
Model Training: Train YOLO for car detection and integrate with RNN or transformer models for trajectory prediction.
Model Validation: Test the integrated model with video inputs to ensure accuracy and efficiency.
Robotics Tasks:
Camera System: Develop and test a rotating camera system that can follow the target car.
Actuator Design: Prototype actuators for proposed actions, such as following or shooting.
System Integration: Integrate the camera and actuator systems with the deep learning model for a cohesive and functional robot.
Testing and Iteration:
System Testing: Test the complete system in controlled environments to ensure it works as intended.
Iterative Improvement: Improve the system based on testing feedback, focusing on accuracy, reliability, and safety.
Final Considerations:
System Safety: Ensure all actions performed by the robot are safe and non-lethal.
Real-Time Performance: Focus on optimizing the system for real-time detection, tracking, and action execution.
Scalability: Design the system to be scalable for different environments and scenarios.
Summary:
Develop a deep learning model using YOLO and trajectory prediction techniques to identify and track a specific car.
Design and implement actuators for proposed actions, ensuring they are integrated with the camera system.
Focus on real-time performance, safety, and scalability to create an effective police chasing robot.
Last updated