Advanced

eBike Radars for Increased Safety

Lindquist, Rebecka LU and Thoft, Andreas LU (2020) BMEM01 20201
Department of Biomedical Engineering
Abstract
The purpose of this thesis is to develop a blind spot detection system for increased safety and comfort for a biker. The solution is based on power-efficient small A111 radars, developed by Acconeer AB, which have a range up to seven meters while consuming only a few milliwatts.

Sensor evaluation is performed for the intended use and from this, concepts are generated. These are tested to obtain the number of sensors, choice of dielectric lens, the angling and the placement needed for the final prototype. User cases representing the intended use of everyday biking are defined and recorded with the final prototype. Lastly, algorithms are tested to increase the accuracy and robustness of the blind spot detector.

The final prototype... (More)
The purpose of this thesis is to develop a blind spot detection system for increased safety and comfort for a biker. The solution is based on power-efficient small A111 radars, developed by Acconeer AB, which have a range up to seven meters while consuming only a few milliwatts.

Sensor evaluation is performed for the intended use and from this, concepts are generated. These are tested to obtain the number of sensors, choice of dielectric lens, the angling and the placement needed for the final prototype. User cases representing the intended use of everyday biking are defined and recorded with the final prototype. Lastly, algorithms are tested to increase the accuracy and robustness of the blind spot detector.

The final prototype consists of three A111 sensors placed under the saddle and are angled 30° horizontally apart and with the range up to seven meters. The three user cases were recorded with increasing difficulty, resulting in one training and one test set which was validated through video recording to obtain ground truth. The blind spot detector performance for the range up to seven meters resulted in the accuracy of 89%. The specificity was close to 95% but the sensitivity was low on the most difficult user case which shows limitations mostly in the far range. This result was obtained using the associated sparse sensor processing with added CFAR and correlation of frames for increased robustness.

The results show that the A111 radar sensor can be used for a blind spot detection system but with a somewhat shorter range than tested. More testing is required to improve the detectability and range further through the optimization of both the prototype and the proposed algorithms. (Less)
Popular Abstract (Swedish)
Radar - Dina ögon i nacken när du cyklar

Cyklister är en utsatt grupp av trafikanter för olyckor. För att öka säkerheten kan radarteknologi användas för övervakning av döda vinkeln samt ge mervärde till fordonet. Med en enkel notis så kommer man undvika risken med att inte veta vad som kommer bakifrån.
Please use this url to cite or link to this publication:
author
Lindquist, Rebecka LU and Thoft, Andreas LU
supervisor
organization
course
BMEM01 20201
year
type
H2 - Master's Degree (Two Years)
subject
keywords
eBikes, radars, safety systems, blind spot, algorithm development
language
English
additional info
2020-04
id
9018873
date added to LUP
2020-06-16 13:47:44
date last changed
2020-06-16 13:47:44
@misc{9018873,
  abstract     = {The purpose of this thesis is to develop a blind spot detection system for increased safety and comfort for a biker. The solution is based on power-efficient small A111 radars, developed by Acconeer AB, which have a range up to seven meters while consuming only a few milliwatts.

Sensor evaluation is performed for the intended use and from this, concepts are generated. These are tested to obtain the number of sensors, choice of dielectric lens, the angling and the placement needed for the final prototype. User cases representing the intended use of everyday biking are defined and recorded with the final prototype. Lastly, algorithms are tested to increase the accuracy and robustness of the blind spot detector.

The final prototype consists of three A111 sensors placed under the saddle and are angled 30° horizontally apart and with the range up to seven meters. The three user cases were recorded with increasing difficulty, resulting in one training and one test set which was validated through video recording to obtain ground truth. The blind spot detector performance for the range up to seven meters resulted in the accuracy of 89%. The specificity was close to 95% but the sensitivity was low on the most difficult user case which shows limitations mostly in the far range. This result was obtained using the associated sparse sensor processing with added CFAR and correlation of frames for increased robustness.

The results show that the A111 radar sensor can be used for a blind spot detection system but with a somewhat shorter range than tested. More testing is required to improve the detectability and range further through the optimization of both the prototype and the proposed algorithms.},
  author       = {Lindquist, Rebecka and Thoft, Andreas},
  keyword      = {eBikes,radars,safety systems,blind spot,algorithm development},
  language     = {eng},
  note         = {Student Paper},
  title        = {eBike Radars for Increased Safety},
  year         = {2020},
}