Skip to content

About the Driverless Kart Project

What is Formula Student Driverless?

Formula Student Driverless (FSD) is a category within the Formula Student competition where teams must develop a vehicle capable of driving autonomously, using only sensors, algorithms, and automatic control—without a driver.

Competition Characteristics

The circuit is delimited by cones that define the track boundaries. Teams don't know the circuit layout in advance, so the vehicle must:

  • Detect the cones using perception systems
  • Infer track boundaries from sensory information
  • Plan a safe trajectory that optimizes speed and stability
  • Execute maneuvers (acceleration, braking, turns) without human intervention

Competition Tests

Tests include autonomous versions of classic Formula Student disciplines:

  • Acceleration: Maximum speed straight line
  • Skid-pad: Figure-8 track to evaluate lateral handling
  • Autocross: One complete lap on the unknown circuit
  • Trackdrive: Multiple laps following the track consistently

Technical Challenges

Developing a Driverless vehicle involves solving multiple engineering challenges:

Perception - Cone detection and classification (blue, yellow, orange) - Environment segmentation and noise filtering - Multi-sensor data fusion

Localization and Estimation - Vehicle state estimation (position, velocity, orientation) - Real-time local map construction - Correction using inertial sensors (IMU) and other systems

Trajectory Planning - Generation of viable paths within detected boundaries - Optimization considering vehicle dynamics - Real-time adaptation to obstacles or incorrect detections

Control and Actuation - Longitudinal control (throttle/braking) - Lateral control (steering) - System coordination to follow the planned trajectory

Safety - Emergency stop systems - Sensor and software integrity validation - Redundancies and fail-safes

Technologies Employed

Teams typically use a combination of:

  • Sensors: Cameras, LIDAR, inertial sensors (IMU), GNSS/INS
  • Software: Robotics frameworks (ROS), computer vision, SLAM algorithms
  • Actuators: Steer-by-wire systems, brake-by-wire, electronic throttle control

What Our Driverless Section Does

The Driverless section of U-Motorsport URJC is responsible for designing and implementing the complete autonomous driving system. Our goal is to develop these systems on the kart as a testing platform and subsequently integrate them into the main competition single-seater.

1. Perception and Sensing

Objective: Detect the environment and extract relevant information for navigation.

  • Cone and track boundary detection using cameras and image processing (computer vision)
  • Use of inertial sensors (IMU) to estimate orientation, accelerations, and vehicle attitude
  • Sensor data fusion for robustness against noise or individual sensor failures

2. Localization and State Estimation

Objective: Know where the vehicle is and how it's moving.

  • Estimation of relative position within the track (where each track boundary is)
  • Dynamic local map registration without relying on a prior map (SLAM)
  • Sensor correction to maintain precision over time

3. Trajectory Planning

Objective: Decide where the vehicle should drive.

  • Generation of a viable path within detected boundaries
  • Optimization of parameters like minimum curvature, smoothness, and safety
  • Real-time adaptation to obstacles or poorly detected cones

4. Control and Actuation

Objective: Make the vehicle follow the planned trajectory.

Longitudinal control - Throttle and braking management to reach and maintain target speed - Response to track changes (braking before curves, acceleration on straights)

Lateral control - Turn execution to follow optimal trajectory - Compensation for slipping and disturbances

Physical actuators - Motorized steering system (steer-by-wire) - Emergency braking system - Electronic throttle control

5. Safety and Redundancies

Objective: Ensure safe operation at all times.

  • Emergency braking mechanism (brake fail-safe) to stop the vehicle if automatic decisions fail
  • Integrity validations of sensors and software
  • Real-time monitoring systems
  • Telemetry for post-execution analysis

6. Hardware-Software Integration and Physical Testing

Objective: Validate the entire system under real conditions.

  • Adaptation of electronic architecture to the real vehicle
  • Testing and calibration on the kart as a testing platform
  • Real circuit test iterations:
  • Cone tracks on campus
  • Slalom courses
  • Unknown curves
  • Increasing speed tests

Previous Season (2024-2025)

During the previous season, the Driverless section laid the groundwork for the project:

Software Development in Simulation

  • Complete Python prototype for autonomous driving
  • Development and validation entirely in simulation
  • Testing of perception, planning, and control algorithms in virtual environment

Physical Architecture Design

  • Initial mechanical architecture design for automatic steering
  • Electronic system planning (sensors, actuators, computation)
  • Specification of braking and sensing components

Virtual Validation

  • Tests in simulated environments with different track configurations
  • Algorithm adjustments based on simulation metrics
  • Development of visualization and debugging tools

Improvements and Advances This Season (2025-2026)

This season we've made a significant qualitative leap, moving from simulation to real implementation:

1. Migration to ROS (Robot Operating System)

Why ROS: - Mature ecosystem with proven libraries for mobile robotics - Modularity: Clear separation between perception, planning, control - Development tools: RViz, rqt, rosbag for visualization and debugging - Compatibility with future more complex systems on the single-seater - Active community and extensive documentation

Impact: - Complete migration from monolithic Python stack to ROS node-based architecture - Better integration with commercial sensors and actuators - Facilitates collaboration between different team subsystems

2. Introduction of the Real Kart (HenaKart)

Testing platform advantages: - Physical validation of all subsystems before integrating on the single-seater - Real environment without compromising the competition vehicle - Rapid iterations: Simpler and safer than testing directly on the Formula car - Outdoor testbed for perception and control algorithms - Educational platform for students, professors, and researchers

Addressing a Critical Need:

Several professors from the Robotics department expressed the need for an outdoor mobile robot platform where they could test and validate their algorithms. Until now, autonomous navigation software and robotic experiments were limited to indoor environments.

This kart project solves two problems simultaneously: 1. Provides a testbed for Driverless technology development 2. Offers researchers an outdoor platform that can navigate the university campus

Current status: - Kart fully operational in manual mode - Actuation systems for steering and braking ordered - Cone detection camera mounted and functional

3. Motorized Steering System (Steer-by-Wire)

Functionality: - Allows the kart to turn based on autonomous controller decisions - Integration with lateral control system in the ROS stack - Preservation of manual mode for supervised operation

Development: - Finalization of mechanical and electronic design - Selection of actuators with adequate torque and speed - Implementation of control interface from ROS

4. Emergency Braking System

Purpose: - Critical safety subsystem to stop the vehicle in case of failure - Automatic activation when: - Loss of communication with control system - Detection of dangerous conditions - Operator emergency commands

Characteristics: - System independent from main control - Redundancies to ensure reliable activation - Integration with telemetry for event logging

5. Perception Optimization without LIDAR

Technical decision: - Operation using only cameras and IMU - Reduction of cost, weight, and complexity - Greater technical challenge, but more aligned with available resources

Implemented improvements: - Optimized cone detection algorithms for higher precision - Reduced latency in image processing - Improved camera calibration for different lighting conditions - Robust filtering to eliminate false positives

6. Improved Planning and Control Algorithms

Approach: - Designed for unknown tracks delimited by cones - Minimized latency for real-time response - Robustness to imperfect or missing detections

Advances: - Implementation of adaptive controllers that adjust parameters according to conditions - Predictive models to anticipate vehicle behavior - Continuous validation with real kart data

Project Final Objectives

Short Term (This Season)

  1. Functional autonomous kart capable of:
  2. Detecting cones with high precision
  3. Navigating simple tracks autonomously

  4. Executing emergency braking reliably

  5. ROS architecture validation for migration to the single-seater

  6. Real data collection to train and improve algorithms

Medium Term

  1. Technology transfer from kart to Formula Student single-seater

  2. Participation in competitions of Formula Student Driverless

  3. Research platform available for:

  4. Students developing bachelor's/master's theses
  5. Professors testing algorithms
  6. Researchers validating new techniques

Long Term

  1. Robust development ecosystem documented for future team generations

  2. Open-source contributions to the mobile robotics and autonomous vehicle community

  3. Academic and industrial collaborations based on the developed platform

For more detailed technical information, see the Assembly, Software, and BOM sections.