DIY Robocars KuaiKai Self-driving Test & Racing Challenge Drew To A Satisfactory End
Self-driving technology has brought significant changes and enormous possibilities to future transportation as well as city life. Engineers, startups, developers, research scientists and talents around the world are dedicated to the development of self-driving, however, when a self-driving engineer or hacker finishes their works and needs testing, the expensive hardware and high-maintenance engineering platform hinder the birth of the next possible big breakthrough in self-driving development.
That’s where the event came alive. DIY Robocars KuaiKai Self-driving Test & Racing drew to a satisfactory end on May 27th last week.
This event gathered 29 engineers from 10 countries worldwide. There are engineers from Egypt, PhD from Cardiff University, professors from Nagoya University, students from University of Bath, engineers from SAIC, makers and software engineer from the US etc.. Together they embarked a 7-day journey of self-driving engineering and software test.
DIY Robocars KuaiKai was divided into two categories: full-sized car onsite test & racing and small-sized car onsite test & racing. The full-sized category starts on May 20th, on which 13 engineers met in C-Zone Hackspace and made a brief introduction of themselves. Small-sized category started on May 24th, but most of them came a bit earlier to learn more about the tracks and meanwhile to check what was going on with the full-sized guys. It was nice to get to know each other, who came from different cultures and backgrounds.
Full-sized Car Onsite Test & Racing
The first day would be grouping and getting familiar with the hardware provided onsite, including the full-sized cars. It became hard to decide how many groups there should be and who belonged to which group as the skills varied and the tasks for the test and racing were quite challenging. So on the first day all the participants were busy setting up the hardware and software deployment while letting the grouping to be decided later. On the second day, car hardware setup was done and a squad went to the racing course to collect some data and get a whole picture of the course.
The racing course was in the science and technology industrial park whose roads were more complex than expected because many people were walking around and the road side was parked with cars. There were 16 missions in total for the full-sized car onsite test & racing, from Pedestrian Avoiding, U-turn, Speed Limits, Queueing, Under Construction Detour, to GPS Outrage, Slope Driving, and Automatic Parking etc.. There was meeting at 5 pm on the second day, some passionate discussion, Q&A, still no clue for grouping and teaming up. It was difficult decision with so many tasks in such short time. Maybe none of the groups would complete the whole challenge. Some people suggested, how about being one big group and going for the Ultimate Challenge: AI Driver vs. Human Driver? This suggestion aroused resonation and unanimous approval. This was super challenging but it would be fun. All of them got excited and wrote down the tasks and arrangement for each person or squad.
In the next few days, all engineers worked together, they were concentrated and dedicated, collecting data, building maps, and optimizing strategies, the passion for engineering challenge was rekindled. All these were contributing to the innovation and development of self-driving technology. DIY Robocars KuaiKai engineers tested their ideas and brought out real-world solutions with various approaches. On 25/26 May, after a simple opening ceremony, the intense test & racing began. Roads were cleared and mission props were set up. The two cars went around the course for improvement on the first racing day, hoping to create new records. During the two days, five human drivers volunteered out of the public came onsite to mark the human driver records along the same missions as the AI driver (the self-driving car built by the participating engineers). Some of them were really fast with nice performance. Of course, they didn’t have the challenge and problems that AI driver would meet.
The testing and record updating went on until 10 pm of May 26th, when night fell and lights on. Until the last minute, the engineers were still optimizing their codes and algorithms. Though the AI Driver didn’t beat the human driver (the scores were pretty close) at this event, the spirits itself explained all. They didn’t care about how many spectators were onsite, they didn’t care how much applause they’re received, their spirits of persistency and dedication has brought them victory. The team won anyway.
Small-sized Car Onsite Test & Racing
The small-sized category participants were as passionate and focused as the full-sized ones. They were energetic and communicative. They shared their own knowledge about small-sized cars.
The car provided onsite for test and racing was Donkey Car. But some of the small-sized participants brought their own cars built by themselves. Erich and Chris are brothers. They are authentic makers who enjoy building and making stuff. They brought their own 3D printed donkey car powered by Raspberry Pi, with cool painted frame. Ahmed showed us his laser-cut wood car. It was small and cute, with uniquely designed appearance and structure. They all were thrilled to be in C-Zone Hackspace and the racing also spiced things up.
On May 24th, small-sized category participants all went to the racing tracks to collect data, build maps and train their models. As the racing tracks were within the Big Data Expo hall, which closed rather early at 5 pm. Most of the small-sized participants would go back to C-Zone Hackspace and continued their work till night. It was unbelievable and amazing experience for them as they were totally absorbed in this process and were having fun.
After a full day of test and optimizing on May 25th, finally came the day for racing on May 26th. Eighteen donkey cars were onsite, ready to go wild and out of control on the tracks. It was time to test the codes and algorithms that the participants worked hard on for the past two days. The racing tracks were set up just like any other DIY Rococars event such as the Oakland Monthly Race organized by Chris Anderson.
Some unexpected issues occurred on the racing day. There was a stage next to the small-sized category racing course. Some colorful lightings from the stage flashed across the racing tracks now and then. Sudden environmental changes had influenced the performance of the racing cars. Finally only four cars completed the racing. Participants suggested that next DIY Robocars event should be within C-Zone Hackspace, which allows more flexibility and brought more innovation to the event. Sure thing!
DIY Robocars KuaiKai event ended successfully on May 27th when all full-sized and small-sized participants gathered together for beers, foods and music. There were so many impressive moments that really touched everyone: the moment when the full-sized team finished the optimization on computer and the car drove along the course in the thick darkness to have one last run, the moment when Chris walked around the racing tracks with his hurt toes, the moment when engineers stayed up late to have a last fight the day before the racing.
Although at last AI driver lost, one day we’ll witness the moment when AI driver beats the human driver, marking the real breakthrough and revolution in self-driving technology. DIY Robocars KuaiKai event and what the participants contributed are driving us closer to this vision. No matter who won the prize finally, the experience and moments are what make it counts for DIY Robocars KuaiKai. Engineers came, talked, learned, grew and had fun, they made the event more meaningful. Everyone worked hard with one hundred percent concentration. Together, they’ll form a powerful open source community for self-driving development and innovation.
Outcomes and Participants
- Rohan, an Electrical Engineering and Robotics student at the Indian Institute of Technology (IIT) Madras, former intern in NVIDIA’s self driving hardware team, co-founder of Dynamove, participant of DIY Robocars KuaiKai and a hackathon organized by PIX in March, has contributed a blog regarding the approaches, outcomes and processes of this event, with more technical details and descriptions. Check here https://wowelec.wordpress.com/2018/05/31/diy-robocars-kuaikai/
- Certain codes, algorithms and programs from the event will be uploaded to community Github https://github.com/pixmoving-moveit
- Final Scores for the full-sized category racing: AI Driver vs. Human Driver 105 : 150 We look forward to moment when AI driver surpasses Human driver in future events.
Last but not least, let’s see what the engineers said on the final two days or on their way back home
PIX is a self-driving startup, building mobility-as-a-service space powered by self-driving technology. PIX realizes the importance of testing for self-driving, which puts safety and daily applications at the first priority. PIX organized DIY Robocars KuaiKai event, which provided car platforms equipped with computing abilities, sensors, by-wire control and other hardware support, also closed urban roads to test and race for global self-driving engineers, who can stay focused on algorithms/software design, optimization and development without worrying about car/hardware to test.There will be future events to further drive the development of self-driving technology in the way of open innovation.