THE DEVELOPMENT OF F1 INSIGHTS, POWERED BY AWS
Start Analysis
F1 Chief Technical Engineer Rob Smedley helps break down each of the new F1 Insights powered by AWS in a series of blogs during the 2021 FORMULA 1 season
The start, or launch as it's known within F1, is absolutely edge of the seat stuff for fans, drivers, and the engineers alike. It is one of the most exciting parts of a Grand Prix race and yet it's all over and done within seconds.
Given the importance of these initial few seconds of a Grand Prix race, the teams put a huge amount of effort into every detail. There are a number of precise and concurrent actions that the driver needs to carry out in harmony with the vehicle control system and power unit. This is by no means a simple task and requires, as is the norm in Formula 1, the optimal blend of both engineering and technical excellence and driving skills.
A simple example: too little throttle in the initial getaway phase will result in the engine bogging and the car pulling away at the slower acceleration than is optimum. On the other hand, too much throttle and the rear wheels will spin excessively. In turn, this also effects the optimum acceleration, and the driver therefore must judge this to perfection. Another example would be the clutch drop phase, here the driver has to manage the clutch position from the moment the lights go out to the point that they have fully engaged it.
There are a lot of things going on in a short amount of time and huge stakes at play. Imagine what the drivers and teams would give to qualify 3 places higher on the grid on a Saturday afternoon? A good start can give you exactly that effect. To understand the F1 launch we are introducing the next of the F1 Insights powered by AWS graphic – Start Analysis.
The aim of Start Analysis is to give the fans a more detailed view of how each driver is able to exploit performance (or not!) in the start or launch phase. In order to understand the graphic, we need to break down the launch into the constituent parts. These are as follows:
- Reaction time – the time from when the start lights go out to when the driver reacts by dropping the hand clutch to the mid (semi engaged/semi slipping) position.
- Initial acceleration phase – how the driver handle the wheelslip, the throttle position, fully releasing the clutch paddle to full engagement, etc.
- How fast are the car and driver in the latter part of the launch – using the slip stream, the overall drag of the car, holding position (making as few changes of direction as possible), etc.
If we take each of the elements above, then it is possible to divide these into phases and graphically output each of these to see how one driver compares to another or what the overall rank is. The two parts of the latest graphic therefore are reaction time, 0-100kph to give an appraisal of the initial start phase performance and finally 0-200kph, which highlights the overall performance from the split second the car starts to move until 200kph.
From these two parts that build the overall graphic we are able to see who reacts the fastest, who controls the critical first phase with the throttle and clutch synchronisation and who has the fastest overall launch up to 200kph.
Start Analysis Modelling
As noted, the outputs of the model are:
- Reaction time
- 0-100 Km/h time and 0-200Km/h time
In the F1 space there are different data sources that we can tap into.
There is the telemetry data that we extract from the cars. In addition, we use the timing data coming from the F1 system position around the track. The transponder (sensor on the cars), triggers the specific loop receiver when it passes it, telling it the specific car ID and the time of day down to 0.0001s. There are typically 25 timing loops placed around the circuit.
The telemetry signals for a single car are synchronised to the clock (or time) within the electronic control unit (ECU). If we take all 20 cars however, there is no single reference clock and therefore each car can have a slightly different time of day against an absolute reference given that they are not synchronised. It should be noted that usually within a team the two cars of that particular outfit will be manually synchronised in order that the analysts and engineers are able to carry out synchronous analysis of the intra team cars.
The timing system is synchronised to the F1 start light system and therefore this gives us an absolute reference and synchronised time of day for all transponder events in terms of duration after the light went out at the start of the race.
In the car telemetry, we also have signals that tells us when the car has “seen” the loop transponder.
If we fuse these elements together, we have a common event in both the car data set and in the timing data. This can then be used to sync the data sets for all the cars to the timing system and henceforth the moment the start lights went out and the race began.
It should be noted that we need more than one car transponder because the moment in which the car detects the loop transponder is also dependent on the car speed.
We then calculate the synchronisation between car and timing system using all the transponder events available during the first lap and then using each of these we can calculate the reaction time of the driver.
What we obtain with the above methodology is a distribution like the following:
We then calculate the median of the distribution, and we estimate the reaction time.
The other output of this new graphic is the start time from 0 to 100 Km/h and from 0 to 200 Km/h
In order to identify these, the model identifies the instant that there is a peak on longitudinal acceleration (green channel in the plot below) and then measure the duration from this point to the point it takes to arrive to 100 and then 200 Km/h.
This is the overall start performance. We should note that the overall duration described above is longer than just considering from the instant the speed deviates from 0kph. This is due to the fact that prior to the car moving, the inertia needs to be overcome. This method also removes any wheel speed sensor jitter or error around the zero minima. For all intents and purposes then, it should give us a more accurate overall picture of performance.
The total time between the lights going out and the car attaining 200 Km/h is the reaction time plus the 0-200 Km/h time.
Offline output – Not available in the TV live feed but calculated within the model and therefore available for post analysis editorials
From the model we also calculate the distances between the cars themselves and between each car and a reference position in the track. This gives us an even deeper insight into start performance.In particular, we digitise the start/finish straight and we indicate using a colour code for what each car was doing during the various phases of the start.
In the plot below you see the drivers’ bar whose x-axis represents the distance to the pink line after the chicane in Monza and the black line represents the apex of the first corner.
Each bar is highlighted in yellow if the driver is not at full throttle, green when the driver is at full throttle and red if the driver is braking reasonably hard (i.e., more than 1.5g). There is then a simple graphical output to show what each driver is doing and how this is affecting their performance.
Instead, the black lines are there to indicate when the car is in tow from the leading car. If the line is dotted it means that that the tow effect is marginal. The tow effect is defined to the distance of the car in front.
In order to calculate the cars position, we use GPS which gives us more granularity than the timing loops.
F1 and AWS we have produced an F1 Insight graphic that unpacks another part of the Formula 1 weekend. We will be able to go back and really get under the skin of how each driver performed at all the crucial phases of the launch. I am looking forward to being able to show how these incredible athletes can react to the start lights going out. Prepare to be surprised by just how good they are.