Crowd sourcing: Teaching Autonomous Vehicles (AV’s), traffic cultures

Crowd sourcing: Teaching Autonomous Vehicles (AV’s), traffic cultures

The world around us is moving fast and many argue that we simply need more time to catch up, longing for ways to save time. No matter what you do or how far your work is, you spend a considerable time driving, stuck in the traffic, or looking for a parking space (Willingham, 2019). All these wasted minutes could have been saved, if your car was self-driving (Worland, 2016). We have been all waiting for autonomous vehicles (AVs) for quite some time and according to Canadian Automated Vehicles Center of Excellence (CAVCOE) they are already here (CAVCOE, 2015) so what is stopping car makers such as Audi which claim their cars to be more than just cars (Audi,2020; Lekach, 2019), from making this long lost human’s dream real? According to Society of Automotive Engineers (SAE), we are far from a fully autonomous vehicle being stuck somewhere between Partial automation and Conditional automation (Business Insider, 2019; Seeker, 2019). Not only dose the autonomous cars need to communicate with the infrastructure and the other cars around them, but, they should fully understand their surroundings as well, being able to effectively communicate with them. To ensure safety, AV’s must be trained to both understand and react, based on traffic cultures which is “easier said than done” (CNBC, 2020) due to their variation form one place to another (Sigporsson & Einarsson, 2019; Cale’, 2011). In search for a way to train AV’s, we came across crowdsourcing (an open innovation method) which can be used in different forms suitable for complex problems in need of creativity (Heidari et al., 2012). The framework concludes practical crowdsourcing methods capable of data gathering in diverse parts of the world to help both CAVCOE and its partner Audi to overcome the barrier of communication. Some of these ways are already in use but with different focus, testing sensors, cars reaction to movements (Audi, 2020), and entertainment (Lekach, 2019). The framework proposes this mixture to gather factors, deeply carved in local cultures, effecting how people behave in traffic. For example, rules and regulations (both written and non-written, wildly accepted in the society), education quality (related to what they learned as they grownup including moral, and ethical matters), how effective is enforcement (strictness of law and how you are treated as the consequence), and finally the road system design (Sigporsson & Einarsson, 2019; Cale’, 2011). Automatic cars are being tested around the world but there has been number of accidents, making the idea of training the Artificial Intelligence (AI) the priority (CNBC, 2020). The four crowdsourcing methods, categorized base on the system needed and the type of data (local traffic cultures) they provide, will help decision makers at Audi and CAVCOE, to ensure AV’s safety while interacting with humans.

No alt text provided for this image

To support our suggestion, let us go to detail about the outcome of each grouping. The two top methods would work nicely while gathering people’s reaction to complicated situations caused by non-written rules (widely accepted by the locals) or moral and ethical matters (learned while growing up in that location). For example, using Captcha (Captcha.net) can be a good option to gather answers to some short questions about traffic culture while a virtual reality games could analyze player’s reaction when put in a complex situation (Onyesolu and Eze, 2011). However, since the user could lie or react differently in a game environment, reliability of data could be questionable. On the other hand, the two bottom methods basically uses sensors to gather real time data which is very reliable data is gathered on enforcement effectiveness and road system design, being able to analyze traffic behaviors in simple situations (Audi Canada, 2020). These four methods can also be categorized based on the type of system they use to gather data which shows two groups of simple supporting and complicated main system. The latter is already available by the partnership Audi has with Disney, used in entertainment area (Lekach, 2019). Another factor is the number of data that could be gathered around the world and the limitation of data gathering, putting Stop spam on top, fallowed by smartphone crowdsourcing (Chatzimilioudis et al., 2012), based on high quantity of data due to the lowest limitation. The third is gathering data from smart car (Audi) providing less quantity of data caused by limited people driving an automated car. The virtual reality might be interesting to a limited group of people around the world despite its potential to put the driver in diverse situations and test the reactions risk free (Onyesolu and Eze, 2011). To summarize using all methods could overcome some limitations, guaranteeing a great quantity of data while ensuring data’s reliability (Xintong et al., 2014).The database can be organized using locations to summarize the traffic cultures related to rules (both written and unwritten), effectiveness of education (type of upbringing or morals and ethics), the strength of rules, and road system designs. The database enables the CAVCOE and its partner Audi to train AI, understanding diverse situations and the reactions needed based on each location. Finally, by having a clear guide map the AVs can predict the reactions and act on them, becoming an integrated part of the traffic system. Let us not forget that we have a long way to go before we get there, rules and regulations must be in place, infrastructures must be ready, even the whether conditions should be considered. It is better to be ready rather than rush into things. It is better to “assume the worst and design for that”, not relying on the technology, promising safety. This is something needed to be done with the community and society considering all safety measures possible (CNBC, 2020).

Audi, 2020, Audi A8 adaptive cruise assist: Attention surplus is in order, https://www.audica.com/technology/intelligence/adaptive-cruise-assist

Business Insider, 2019, Why don’t we have self-driving cars yet? YouTube, https://youtu.be/qf6VrDZ04EQ 

CAVOE. 2020. https://www.cavcoe.com/

CNBC, 2020, Why don’t we have self-driving cars yet? https://youtu.be/qf6VrDZ04EQ

Chatzimilioudis, Konstantinidis and Laoudias, Crowdsourcing with Smartphones, University of Cyprus, file:///C:/Users/Hosna/Downloads/IC_ICSI-2012-01-0015.R2_Chatzimilioudis-NEW%20(1).pdf

Heidari, Akhavannia, and Kannangara, 2012, To Internationalize Rapidly from Inception: Crowdsource, TIM Review, https://timreview.ca/sites/default/files/article_PDF/Heidari_et_al_TIMReview_October2012.pdf

Lekach, 2019, Audi partners with Disney to entertain car passengers with VR, Mashable, https://mashable.com/article/audi-disney-vr-holoride-ces-2019/

Onyesolu and Eze, 2011, Understanding virtual reality technology: Advances and Applications, Computer Science, https://pdfs.semanticscholar.org/765d/fb3e35aa6ac9cac49167312000befb5ed0e5.pdf?_ga=2.28960892.1365028458. 1593045451-504970682.1592676390

Sigporsson and Einarsson, 2019, Reykjavik University, https://www.ru.is/media/tvd/skjol/HS.pdf

Willingham, 2019, Commuters waste an average of 54 hours a year stalled in traffic, study says, CNN, https://www.cnn.com/2019/08/22/us/traffic-commute-gridlock-transportation-study-trnd/index.html

Worland, 2016, Self-Driving Cars Could Help Save the Environment- or Ruin it. It depends on Us, TIME, https://time.com/4476614/self-driving-cars-environment

Xintong, Hongzhi, Song, and Hong, 2014, Brief survey of crowdsourcing for data mining, Expert system with applications, https://www.sciencedirect.com/science/article/abs/pii/S0957417414003984


要查看或添加评论,请登录

社区洞察

其他会员也浏览了