The candidate will join Machine Intelligence team at ACSL. This team works on various functions of the drone running on the onboard computer, including:
• Computer vision and robotics related technologies, e.g. SLAM, obstacle detection and avoidance.
• AI, for instance human detection or safe landing detection.
• Security, for instance data encryption and securing the drone from intrusion.
• Integration with peripherals, for instance FPV and inspection cameras.
The candidate will work closely with members working on computer vision and robotics related technologies. The team members come from all around the world, and depending on the member, we use both English and Japanese to communicate.

The main responsibility for this role is to improve the performance and to develop new algorithms with respect to drone collision and obstacle avoidance. A crucial aspect of flying the drone is safety. Whenever a pilot flies the drone far away, it becomes impossible to determine how close the drone is to objects around it. It is therefore vital to drone operation to rely on the drone flying around obstacles or stopping before hitting them.
The basis for these algorithms is depth information from the onboard sensors. This information can be noisy – both in the gaussian sense, and that the drone might hallucinate things that are not there. Any algorithm based on this data therefore needs to be robust to these phenomena. In addition, any algorithm needs to run real-time at the edge on a compute-restricted platform.
Since collision and obstacle avoidance is closely related to drone control, the candidate is expected to work closely with the Flight Controls team, which develop flight algorithms for the drone. The flight controller is running on a platform separate to the collision and obstacle avoidance algorithms, meaning that effective communication over CAN/UART is essential for a reliable system.
The development partially happens in simulation, but will increasingly need to get tested on a real drone. As such, the candidate is expected to visit the ACSL drone test field for real-world testing.
The ACSL code base is written in C++ and CUDA/OpenCL. As such, the candidate is expected to have good knowledge of C++. In addition, we use tools and processes like CMake, unit testing (GoogleTest), git (GitLab and GitLab CI) and Docker, so experience in any of these is a plus.

Must:
• Background in robotics, computer vision or machine learning.
• Solid knowledge of C++ and related tools.
• Effective communication skills.
• Passion for improving and perfecting ACSL products.
• Have a good habit of documenting code and processes.
Nice-to-have:
• Experience with scripting languages (Python, Bash).
• Conversational English, for native Japanese speakers, or conversational Japanese for non-native speak.

Hulic Kasai Rinkai Building, 3-6-4 Rinkaicho, Edogawa-ku, Tokyo134-0086, Japan

Experience and ability will be considered in accordance with the company's rules, and will be decided after interviews.

Full-time employee

09:00 - 18:00 (flex time)

128 days a year, Saturdays, Sundays and holidays
Paid leave: 10 days (Varies depending on month of hire)

Social insurance, health insurance, employees' pension, unemployment insurance, and workers' compensation insurance, self-development support, and nearby housing allowance