The goal of this proposal is to have two or three RC-Cars/Autonomous robots drive around autonomously and collect data of all kinds of information belonging to each car or the track condition and make this information available to receive/buy for every other car over the Streamr (Core) Network (PubSub). Each car will be able to
receive/buy and transmit/sell data.
The use case is already described here. There is a tremendous amount of data produced by every single car each day which is just vanishing without any usage. Besides cars, autonomous robots (e.g. last mile delivery robots) could also profit from a reliable message exchange mechanism to decide about the fastes/safest route to take.
Possible scenarios within this PoC
The following data could be shared/sold between cars:
- Tank station price
- Waiting time during traffic light (to create a fair price for advertisements next to that traffic
- Stop sign or traffic jam → route enhancement (route condition data)
- Free parking slots to reduce traffic jam through slow driving cars (looking to find a
sparking spot), information from parking sensor or other cars
- Own speed/average speed and odometry
- Own route planning and location
- Time to complete one round of the track
- Distance traveled (Insurance could be paid on a seconds basis)
- Battery/Fuel usage
- Parking duration estimation of other cars (from license plate as specific car ID)
Most data can be sent/bought car2car, but some information like waiting time in front of a
traffic light could also be sold over the marketplace to advertising companies to create a fair
value of the add according to the quantity of cars waiting and according to the average
Also the duration of cars parking at the specific parking slots is data useful for outside
(Estimated level of implementation difficulty: 1 easy to 3 hard)
(i) Create a possible track (1)
(ii) Assemble one car and make it drive autonomously around the track (Path planning or/and end-
to-end learning) (2)
(iii) Gather “easy to get” data with that specific car and share it over the streamr network (e.g.
distance traveled, time to complete one track, speed/average speed, battery usage) (1)
(iv) Create a simple dashboard to display all available gathered data and sensor states
(v) Add more sophisticated algorithms to detect stop sign/barrier/traffic light/parking slots one after
another, starting with obstacle detection to enhance route planning and avoid traffic jams (3)
To test the feasibility of the concept, I would start out with a single car, make it drive around autonomously and collect useful data according to the roadmap/scenarios above.
Main parts for a single car (rather sophisticated, partly based on https://github.com/mit-racecar/hardware and https://github.com/NVIDIA-AI-IOT/jetracer/blob/master/docs/tamiya/bill_of_materials.md ):
This is for the main car and the heavy Computer Vision (CV) parts. Additional cars could be build according to the popular Donkey Car https://www.donkeycar.com/ which would then consume/subscribe to the data gathered by the main car.
Cars and scenarios from above could be added one by one, if the project is successful.
Background of developing