Train AI racers fast and free — reinforcement learning you can actually teach

A 1/28 car brings the racetrack into the classroom: unlimited training on your own computer, and everyone gets time on a real track.

Reinforcement learning is a loop

You do not write the model. You run it, look at what happened, change something, and run it again.

  1. Design

    Decide what the car is allowed to do, and what counts as doing it well. The action space and the reward function are the only instructions you can give a model, which makes this where the loop really begins.

    The track laid out as a mesh of numbered segments, with the car drawn at each position along the lap.
    Designing starts from the track itself — the whole course laid out segment by segment, with the car at every point along the lap.

    Related products: TopRacer Console

  2. Train

    Try, fail, and try again in the 3D simulator, with PPO learning from each attempt. Crashing costs nothing, and it all runs on your own machine.

    The Console mid-training: a mean-reward chart climbing across iterations, the simulator's 3D camera view, and the car's attempts traced in red on the track map.
    PPO training live in the Console — mean reward per iteration, the simulator camera, and every attempt drawn on the track.

    Related products: TopRacer Console

  3. Deploy and drive it

    Send the trained model to a TopRacer Car and let it drive TopRacer Map 2018 for real. Nothing has to be rewritten to make the trip.

    A classroom test run: students and their teacher around the TopRacer track on the floor, the car mid-lap. Faces blurred for privacy.
    The model on a real car in a real classroom — a lap on TopRacer Map 2018 with the whole class watching.

    Related products: TopRacer Map 2018 TopRacer Car MiniDeepCar Online race hosting

  4. Evaluate and think

    Read the lap times, the driving path, and the mistakes to work out what actually happened, then decide what to change. That thinking is worth more than another round of training.

    The run viewer replaying a recorded drive: the camera frame at one moment of the run, playback controls, and a chart of the steering the model commanded frame by frame.
    One real drive replayed frame by frame, with the steering the model actually commanded.

    Related products: TopRacer Map 2018 TopRacer Car MiniDeepCar

Back to step one The new settings go back to step one and train another round. The loop runs as long as you want it to, and every round you can see how much better it got.

Who TopRacer is for

We chose a 1/28 car so that every student gets a turn on a real one — instead of a class watching one big car perform.

For teachers

One classroom, one teacher, the whole AI loop in one semester

The 1/18 big-car packages are built for big events: a big car needs a big track, the big track needs a gymnasium, rent, air conditioning and a crew to move it — so it happens once a year.

TopRacer uses a 1/28 car: the track unrolls on the classroom floor and rolls up into a cabinet after class. The training software is free — paid training platforms are hard to even budget for; free needs no sign-off at all. Training runs on the students' own computers with no cloud bill, so the total budget is final the day you order.

Most of all, the lessons stay coherent: on some platforms one training run takes hours — press train this class, see results next week, thread lost. TopRacer finishes a round on a laptop in minutes: change the reward, watch the result, discuss it, all in one period. And from first lesson to the class race, one teacher runs the whole thing, with every student's records in the tool for grading.

Tell us what you need

For students

Change the hypothesis, run the experiment — then prove it on the track

The training system is free, which means there is no cap on training runs. A tweaked reward function is a hypothesis; pressing train is an experiment. Guessed wrong? Change it and run again — ten more times tonight if you like.

On other platforms every experiment burns money. Yours are unlimited and free — which is what the scientific method is supposed to feel like: hypothesise, test, repeat.

Then comes the real race: the car fits on a desk, the track sits at the back of the room. The model you built experiment by experiment deploys on the spot — lap times go on the leaderboard, and you race your class this term.

Download TopRacer Console for free

For event organizers

Same venue, same budget, ten times the show

What did AI racing events used to look like? One track, one big car, one act: cars taking turns on the same course. Fresh the first year, a rerun the second, phones out by the third. The budget was never the problem — once the big car eats the whole venue, there is nothing left to vary.

At 1/28 the same floor becomes a different place: complex courses fit — the track library runs from short sprints to hundred-metre layouts — and several different tracks run side by side. Qualifiers, knockouts, parallel heats: change the format every year.

The race system keeps up with the creativity: entries, lap times, trajectories and the leaderboard are automatic, ready for the big screen. Competitors practise at home for free before they arrive, so the final is worth watching.

Host a race

Schools and groups

TopRacer Car and TopRacer Map 2018 are coming soon. If you are evaluating them for a course or a club, tell us what you need.