Artificial Intelligence in Mountainbiking to optimize rider performance

Supervisor

Assoc Prof Dr Hermann Klug (hermann.klug@plus.ac.at)

Short description

Mountainbiking is a complex sport. Analogue to Formula 1 car racing, many spatial and temporal parameters can be captured with sensors, either mounted on a bike (speed, cadence, tire coat air pressure, gear used, set post position up/down, pedalling power right/left, brake use front/rear, gear shifts) or the athlete's body (wearables like heart rate chest belt, gyration sensor). Additionally, terrain information classified in coordinates (xyz) and underground conditions/material (grass, soil, stones, roots, wetness) are available besides general weather conditions (wind, rain, temperature, darkness, shadow, sun). A particular trail section of the cross country race track in Koppl close to Salzburg will be used as a test bed. Performant riders will ride the pre-defined track numerous times with equipped sensors. Many measurements per seconds of riding time contribute to a big data package for many rides to be analysed with data science methods. The objective is to analyse similarities of riding performance of single and/or many riders and to discover likely speed decrease at certain sub sections of the track. While the sense of place is most important, cross sensor parameter analysis is important, too. For instance, the pedalling power (wattmeter) and speed (cadence sensor) on a slope in relation to the location of a gear switch and heart rate. Are there areas of increasing slope, where the gearshift is happening too late and thus a speed reduction occurs? The ultimate objective of the algorithm(s) to be developed is to improve/increase riding speed through place-based indications of performance measures (power on pedal, cadence, gearshift).

 

References & datasets

• LiDAR DEM and DSM data from the Federal State of Salzburg in 1 m resolution
• Probably a very high resolution DEM taken from a drone flight campaign
• Race track including field measurements from underground conditions on the track
• Measurements from sensors bike and athlete's sensor during the riding exercise

Materials and methods

Materials and data as descried above will be provided. The algorithm to be developed should learn from the exercises and contextual inputs

Prerequisites/qualification

Interest in mountainbiking, solid skills in spatial analysis, programming and automation of processes, use of artificial intelligence (AI)

Planned Start

Summer 2022