Autonomous Platform Generation 4

Autonomous Platform (Generation 4) is a platform at Infotiv on which internal and external research projects can be tested — particularly autonomous driving algorithms. The system features a modular mono-repository collecting all software and hardware design components in one place.

Autonomous Navigation with LiDAR and SLAM

Imitation Learning

Raw footage:

System Architecture

The platform has three software layers:

• High-level control — autonomous driving algorithms, Gazebo digital twin, RViz
• Low-level control — Raspberry Pi 4b, ROS2 \cmd_vel to CAN bus bridge
• Embedded control — ECU firmware (Arduino/PlatformIO) for steering and propulsion (SPCU)

Capabilities

• Drive-by-wire with Xbox 360 controller
• Imitation Learning via Behavioral Cloning (BC) and Human Gated Dataset Aggregation (HG-DAgger)
• LiDAR-based SLAM and autonomous navigation with Nav2
• Gazebo digital twin synchronized with the physical platform
• Fully Dockerized software stack

Master Theses

• Spring 2023 (Fredrik Juthe, Erik Magnusson): E/E and software architecture design
• Spring 2024 (Arvid Petersén, Johan Wellander): Imitation learning pipeline (BC + HG-DAgger)
• Spring 2025 (David Espedalen, Anton Stigemyr Hill): LiDAR-based SLAM and autonomous navigation

All three theses are supervised by Hamid Ebadi at Infotiv AB.

Links

Autonomous Platform GitHub repository