Hybrid Autonomous Connected Vehicle platooning with Federated Learning: State of the art and simulation Framework

Actas de las XV Jornadas de Ingeniería Telemática (JITEL 2021), A Coruña (España), 27-29 de octubre de 2021. Hybrid Autonomous Connected Vehicle platooning with Federated Learning: State of the art and simulation Framework Prashanth Kannan, Yaqoob Al-Zuhairi, Mónica Aguilar Igartua Department of Network Engineering Universitat Politècnica de Catalunya (UPC) Barcelona, Spain [email protected], yaqoob.al-zuhai[email protected], [email protected] In this paper, we present a brief overview about technologies and state of art methods being used for Hybrid Autonomous Connected Vehicles (HACV). Moreover, Federated Learning (FL) help in avoiding transmission of raw local data in the design of machine learning models for diverse purposes, which help to ensure privacy of sensible data. On the other hand, to reduce tailpipe emissions hybrid electric vehicles are required, as complete conversion of engines to electric might take time. Communication and connectivity of vehicles and infrastructure are increasing for connected autonomous vehicles (CAV) that play a vital role in future transportation. High speed, reliable and efficient communication between vehicles and infrastructure is made possible with fifth generation (5G) wireless technology. Here, Platooning and 5G helps in joining a cluster of vehicles that aids in reducing fuel consumption and allowing exchange of energy using wireless charging. Keywords: Platooning, FL, VANET, V2X, Autonomous vehicles 1. INTRODUCTION The purpose/goal of this article is to address the gap in utilizing diverse wireless technologies for vehicular networks and creating a new federated learning architecture for protection of data. In this scenario platooning of vehicles is integrated with vehicular communication technology using either 5G/Advance long term evolution (LTE) or IEEE 802.11p. This framework might help to increase safety and efficiency of vehicles and allows wireless charging of vehicle batteries. We also focus on analyzing the impact of communication failure to platoon vehicles in case of mixed traffic conditions having manned and unmanned vehicles together. The rest of the paper provides a brief outlay of current technology embedded with hybrid vehicles (fuel and electric), connected vehicles (5G and/or IEEE 802.11p), vehicular communication and federated learning (FL), along with different simulators that are useful in developing the scenarios and test our proposals. 1.1. Shared Autonomous and Connected Hybrid Vehicles One of the current major directions of research is intelligent transportation systems (ITS) where cities are connected through autonomous vehicles (AV) and wireless technology that helps in traffic flow management, vehicular communication and decreasing CO2 emissions to avoid the ‘urban road flow congestion’, reduce ‘environmental pollution’ and ensure ‘safety and ride comfort’ of the vehicles. The level of autonomy has been divided into six levels based on Society of Automotive Engineers (SAE J3016) standards, ranging from Level 0 to Level 5 [1]. According to the survey [2], 69% of respondents stated that fully automated driving will reach a 50% market share before 2050. However, understanding preferences and public response about AVs could result in Shared Autonomous Vehicles (SAV) and Connected Autonomous Vehicles (CAV) that are closely linked to ITS. SAV and CAV are beneficial when compared to that of AV [3]. It seems clear that SAVs are an interesting alternate mode of transport that replaces private vehicles [4]. However, uncertainties of utilizing SAV and CAV with respect to real world traffic made it a challenging aspect when considering mixed traffic scenarios. Also, the powertrain aspect is to be considered as a major area of research that needs to be focused, SAV and CAV with hybrid electric mode have a great potential in the depletion of tailpipe emissions and energy utilization [5]. At present, zero emission from vehicles is possible to a limited extent as complete conversion of conventional internal combustion vehicle to electric vehicle might take a lot of time due to lack of charging infrastructure, range of electric vehicle, charging time and also initial cost of the vehicle [6]. Due to the stringent emission norms and also with the strict regulations made a change to look for alternate 277 power sources. By 2050, European Union aims to reduce greenhouse gas emissions to zero that helps in development of research activities [7]. Hence, to reduce the percentage of CO2 emissions hybrid electric autonomous vehicles (HEAV) play a vital role. This reduces fuel consumption by 40-50% by increasing the efficiency [8]. 1.2. Platoon vehicles Communication and collaboration of vehicles while in travel is possible using the platoon system [9, 10] which improves vehicle safety by reducing accidents, improving fuel efficiency and increasing driver comfort. Here a group of vehicles platoon by having a platoon leader (PL) vehicle that collects the information of location and road data ahead and leads the platoon members (PM) vehicles that are connected behind to navigate on road. As a result, lateral and longitudinal motion control of PM vehicles could be maintained [11]. However, in critical situations drivers should stay alert to avoid the risk of fatalities while vehicles are connected in platoon [12]. Hence, autonomous platoon vehicles (APV) are playing a vital role to eliminate the human errors when travelling in a mixed traffic flow scenario. This is operated with the help of cooperative adaptive cruise control (CACC) that combines adaptive cruise control (ACC) and vehicle to vehicle communication [13]. APV helps not only to connect commercial vehicles, but also helps to connect SAV by exchanging the vehicle information such as trajectory predictions, kinematic status, sensors and camera data. The length of platoon and number of platoon vehicles are very much crucial in managing safe travel in urban and highways scenarios. 1.3. V2V and V2X communication To avoid fatality, maintain efficiency and increase the safety of SAV and CAV, vehicle to vehicle (V2V) and vehicle to infrastructure (V2X) communications are necessary. Here, road-side units (RSUs) are deployed over traffic sign boards, traffic signals, gas stations, etc., which establishes a wireless communication script [14]. Vehicular networks allow vehicles mounted with communication devices such as on-board units (OBUs) and other sensors that communicate using a dedicated short-range communication (DSRC) based on IEEE 802.11p standard or with cellular V2X (CV2X) [15]. Proper routing protocols developed for vehicular communication must cope with potentially high speeds of the nodes (i.e., vehicles) as well as with variable network connectivity having either sparse and dense scenarios throughout time. Also, they should consider different data priorities, e.g., sending high priority data to avoid collisions, low priority data about traffic congestion information for traffic management purposes [16]. 1.4. LTE/5G in Vehicles Advanced wireless technologies such as LTE and fifth generation (5G) communication plays a vital role in transforming V2V and V2X communication systems. [17] suggests that to ensure the safety of the vehicle, decision making and controlling need to be maintained. Implementing 5G technology along with V2X could sustain error-free navigation and accurate trajectory range than that of 4G or DSRC wireless technologies. For instance, applying the brakes in a panic situation using 4G might be 1.5 m for a vehicle, whereas for 5G it would be around just 2.5 cm. This provides promising performance improvement of braking with 5G technology compared to that of 4G [18]. Also, one of the important areas of research is hybrid communication using LTE/5G combined with DSRC/802.11p. The advantages and features of a hybrid cellular and DSRC technology is considered as an alternative for V2X communication for ITS. This has an ability of large coverage, better performance, and universal deployment when compared with only using LTE/5G or IEEE 802.11p [19, 20]. One of recent studies tests for advanced LTE have been conducted for 50 kmph at different load conditions where carriers implemented 10MHz and 15MHz to analyse the channel quality. The results reflected that carrier attained an average of 17 Mb/s and 30 Mb/s respectively [21]. Fig.1. Typical architecture of HACV Platooning with Federated Learning. 1.5. Federated Learning in AV With the increase in connectivity and communication of vehicles, huge amounts of data is being migrated from one point to another. Data privacy and security are major areas that need to be focused to avoid the leakage of sensible information that are being shared. Mitigation of data leakage is due to inadequate storage and computing for single node points such as vehicle or road side units. Hence, federated learning (FL) plays a vital role in managing privacy of data [22]. FL is one of the machine learning (ML) approaches where models are trained locally with data available internally. Hence, only updates from the model are being shared and direct transmission of data will not be present, thus ensuring privacy and reducing latency. The process of FL can be executed in the following process: (i) Selection of clients; (ii) Dissemination model; (iii) Distributed learning; (iv) Feedback from clients; (v) Aggregation; (vi) Testing model; (vii) Model update. A possible implementation of FL in a vehicular network is using tree-based learning method and convolution neural 278 networking (CNN), which eliminates the presence of malicious data and avoids security failures to maximum extent [23]. However, as the vehicular environment is dynamic and complex, collaboration between vehicles and RSU is a challenging task that creates delay in response. Here, we present the topics that will be studied in the first authors’ doctoral thesis. In section II we highlight some of the simulators that are used to evaluate the performance of the proposed work. Section III summarizes highlights of some related work on connected autonomous and hybrid vehicles, federated learning and platooning. Since we consider the presence of hybrid autonomous electric vehicles in our proposals. Finally, section IV depicts innovative ways in collaborating the vehicles using platooning and effective ways in wireless charging of vehicles while travelling, which will be taken into account in the design of our FL framework. Fig.1. represents the basic architecture of the proposed framework platooning and wireless charging for HACV based on FL. 2. SIMULATION FRAMEWORK To avoid real world congestion and collision with pedestrians, vehicles and other objects, AVs and CAVs are tested using multiple simulators that provide better accuracy and reliable results. Numerous simulators such as SUMO, OMNET++, VENTOS, Veins, Plexe, etc. are today available. A brief overview about main simulators used for experimentation in vehicular communications, is presented below. 2.1. SUMO Simulation of Urban Mobility (SUMO) [24] is an open source, microscopic traffic simulator software and one of the widely used software for simulating the ITS and route planning for a particular location. SUMO also follows the physics and dynamics of various types and quantities of vehicles [24]. Generation of road network is possible with netconvert, netgenerate and netedit whereas traffic generation is possible with DUArouter, JTRrouter OD2trips, MArouter, and DFrouter. Once vehicle routing and traffic data are classified, communication between vehicles and infrastructure are classified using OMNET++. 2.2. OMNET++ and Veins Objective Modular Network Testbed in C++ (OMNET++) [25] is a discrete event simulator with an integrated development environment (IDE) based on C++ language that helps in building network simulations. This consists of a tested and predefined framework to determine and simulate various networks, such as INET (TCP/IP stack), Veins/MIXIM (WAVE/DSRC stack), controller area network (CAN) and SimuLTE (LTE cellular network). One among them is Veins [32], which integrated with OMNET++ and SUMO to form a full simulation framework for vehicular networks. Veins includes a road and communication network that helps to simulate VANET applications for mobility vehicles. 2.3. VENTOS AND PLEXE Vehicular Network Open Simulator (VENTOS) [26] is one of collaborative traffic flow analysis software that allows researchers to implement algorithms and protocols about DSRC. It is an open-source simulator that helps the V2X community to work with transportation, vehicular network and control theory research fields. VENTOS architecture has additional features for cooperative adaptive cruise control (CACC) in SUMO through TRACI, that helps in platooning of the vehicles. VENTOS also supports veins for V2X communication with IEEE 802.11p [27]. Another open-source simulator that is similar to VENTOS is PLEXE [28] which is mostly used for Platooning or CACC of the vehicles. As Plexe is an extension of veins framework, this helps in simulating realistic environments, maintaining the physics and dynamics of the vehicles along with an inter-vehicle communication protocol stack. 3. RELATED WORK The following papers address key issues of the research area that need to be focused on: The work [5] implemented longitudinal autonomous driving for hybrid electric vehicles. Along with access for V2X/V2V is made simple using game theory approach for longitudinal autonomous driving control framework. This helps in achieving multiple objectives by coordinating internal combustion engine, electric motor and vehicle braking. The simulation results from the experiments suggest reduction in fuel consumption, improving driving comfort and car-following ability. The authors of [29] developed a V2V communication network to improve a truck platooning that enables sharing of data, e.g., speed of vehicle. Here, two antennas were placed at the front and rear of the PL vehicle that provides good results in a real scenario. In [30] the authors considered an advanced vehicular networking called Federated Vehicular Networking that consists of DSRC and mmWave communication for stable and scalable performance and supports machine learning (ML) and FL. Also, they have tested several routing algorithms using NetworkX, a graphbased simulator. The authors in [31] proposed a new FL algorithm for designing a CAV controller to predict and perform real time decisions. The authors proposed a novel dynamic federated proximal (DFP) algorithm for FL training. Simulation results show that the controller helps in identifying the accurate change over speed in different traffic scenarios and also, helps in developing wireless connectivity of CAVs. 4. ONGOING WORK To improve the energy efficiency of vehicles, maintaining the safety of vehicular data and reducing the CO2 emission are the main areas that are being focused on this thesis. The state of art (SOA) technology that is being implemented in the ongoing work is followed in four stages: (i) First stage is to design and implement a simulation framework for hybrid electric vehicle technology with shared connected autonomous vehicles that helps in maintaining the change in powertrain from engine to electric, based on the flow of traffic, type of road network such as urban area and highway. (ii) In the second stage, communication between vehicles is made for a cluster of vehicles forming a platoon. Implementing HEAV technology with CACC or platooning helps in reducing fuel consumption while vehicles are in platoon. This also helps in improving the state of charge (SOC) of the battery that helps 279 in charging batteries while vehicles are connected in platoons. (iii) The third stage of our ongoing work is to utilize 5G or advanced LTE combined with IEEE 802.11p/DSRC technologies that can safeguard the vehicles from collision and maintain high speeds and proper communication between vehicles without any disturbance. (iv) Finally, to sustain the privacy of data that are being shared between vehicles and RSU, advanced machine learning architectures such as FL will be employed. 5. ACKNOWLEDGEMENTS This work was supported by the Spanish Government under research project “Enhancing Communication Protocols with Machine Learning while Protecting Sensitive Data (COMPROMISE)” PID2020-113795RBC31/AEI/10.13039/501100011033. REFERENCES [1] “Taxonomy and Definitions for Terms Related to Driving Automation Systems for On Road Motor Vehicles”, SAE Standard J3016, 2018. [2] M. Kyriakidis, R. Happee, J.C. de Winter, “Public opinion on automated driving: results of an international questionnaire among 5000 respondents”, Transport Res. F: Traffic Psychol. Behav., vol. 32, pp. 127-140, 2015. [3] Daniela Paddeu, Ioannis Tsouros, Graham Parkhurst, Amalia Polydoropoulou, Ian Shergold, “A study of users preferences after a brief exposure in a Shared Autonomous Vehicle (SAV)”, Transportation Research Procedia, vol. 52, pp. 533-540, 2021, https://doi.org/10.1016/j.trpro.2021.01.063. [4] Tony Z. Zhang, T. Donna Chen, “Smart charging management for shared autonomous electric vehicle fleets: A Puget Sound case study”, Transportation Research Part D: Transport and Environment, vol. 78, 2020, https://doi.org/10.1016/j.trd.2019.11.013. [5] Shuo Cheng, Liang Li, Xiang Chen, Sheng-nan Fang, Xiang-yu Wang, Xiu-heng Wu, Wei-bing Li, “Longitudinal autonomous driving based on game theory for intelligent hybrid electric vehicles with connectivity”, Applied Energy, vol. 268, 2020, https://doi.org/10.1016/j.apenergy.2020.115030. [6] Shaik Amjad, R. Rudramoorthy, P. Sadagopan, S. Neelakrishnan, "Implementation and evaluation of change-over speed in plug-in hybrid electric two wheeler", Energy, vol. 109, 2016, pp. 858-865, https://doi.org/10.1016/j.energy.2016.05.025. [7] Union, E., 2021. “EU Climate Action and the European Green Deal”, Accessed 11-02-2021 2021. https://ec.europa.eu/clima/policies/strategies/2050_en. [8] Farooq Sher, Sihong Chen, Ali Raza, Tahir Rasheed, Omid Razmkhah, Tazien Rashid, Piyya M. Rafi-ul-Shan, Begum Erten, "Novel strategies to reduce engine emissions and improve energy efficiency in hybrid vehicles", Cleaner Engineering and Technology, vol. 2, 2021, https://doi.org/10.1016/j.clet.2021.100074. [9] Carl Bergenheml, Henrik Pettersson, Erik Coelingh, Cristofer Englund, Steven Shladover, Sadayuki Tsugawa, "Overview of platooning systems", 19th ITS World Congress, Vienna, Austria, 22/26 October 2012. [10] Heikoop, D., de Winter, J., van Arem, B., Stanton, N. A., "Effects of platooning on signal-detection performance, workload, and stress: A driving simulator study", Applied Ergonomics: human factors in technology and society, vol. 60, pp. 116-127, 2017, https://doi.org/10.1016/j.apergo.2016.10.016. [11] A. Sarker, C. Qiu, H. Shen, "Quick and autonomous platoon maintenance in vehicle dynamics for distributed vehicle platoon networks", in: Proc. of 2nd International Conference on Internet-of-Things Design and Implementation, pp. 203–208, 2017. [12] Banks, V.A., Stanton, N.A., Harvey, C., "Sub-systems on the road to vehicle automation: hands and feet free but not ‘mind’ free driving", Safety Science, vol. 62, pp. 505-514, 2014, http://dx.doi.org/10.1016/j.ssci.2013.10.014. [13] P.K. Singh, G.S. Tabjul, M. Imran, S.K. Nandi, S. Nandi, "Impact of security attacks on cooperative driving use case: CACC platooning", in: Proc. of TENCON, pp. 0138–0143, 2018. [14] Hong Zhang, Xinxin Lu, "Vehicle communication network in intelligent transportation systems based on Internet of Things", Computer Communications, vol. 160, pp. 799–806, 2020. [15] Michael Lee, Travis Atkison, "VANET applications: Past, present, and future", Vehicular Communications, vol. 28, 2021, https://doi.org/10.1016/j.vehcom.2020.100310. [16] Dahiya, A., Chauhan, R., "A comparative study of MANET and VANET environment", Journal of computing, vol. 2, no. 7, pp. 87-92, 2010. [17] Y. Shi, Q. Han, W. Shen, X. Wang, "A Multi-Layer Collaboration Framework for Industrial Parks with 5G Vehicle-to-Everything Networks", Engineering, 2021, https://doi.org/10.1016/j.eng.2020.12.021 [18] Guevara L, Auat Cheein F. "The role of 5G technologies: challenges in smart cities and intelligent transportation systems", Sustainability, 12(16):6469, 2020, https://doi.org/10.3390/su12166469. [19] G. Araniti, C. Campolo, M. Condoluci, A. Iera and A. Molinaro, "LTE for vehicular networking: a survey", in IEEE Communications Magazine, vol. 51, No. 5, pp. 148-157, May 2013, doi: 10.1109/MCOM.2013.6515060. [20] A. Burbano-Abril, R.M. Edwards, V. Rangel-Licea, R. Aquino-Santos, M. Lopez-Guerrero, R.S. Kalawsky, M. Behjati, "Modeling and analysis of LTE connectivity in a high mobility vehicular environment", Computers & Electrical Engineering, vol. 68, pp. 322336, 2018, https://doi.org/10.1016/j.compeleceng.2018.04.007. [21] S. Avallone, N. Pasquino, S. Zinno and D. Casillo, "Smartphone-based measurements of LTE network performance", 2017 IEEE International Instrumentation and Measurement Technology Conference (I2MTC), pp. 1-6, 2017, doi: 10.1109/I2MTC.2017.7969854. [22] Jakub Konečný, H. Brendan McMahan, Felix X. Yu, Peter Richtárik, Ananda Theertha Suresh, Dave Bacon, "Federated Learning: Strategies for Improving Communication Efficiency", arXiv:1610.05492, 2016. [23] Y. Lu, X. Huang, Y. Dai, S. Maharjan and Y. Zhang, "Federated Learning for Data Privacy Preservation in Vehicular Cyber-Physical Systems", IEEE Network, vol. 34, no. 3, pp. 50-56, May/June 2020, doi: 10.1109/MNET.011.1900317. [24] Daniel Krajzewicz, 2010, "Traffic Simulation with SUMO – Simulation of Urban Mobility", International Series in Operations Research & Management Science, Fundamentals of Traffic Simulation, chapter 0, pp. 269-293, Springer, DOI: 10.1007/978-1-4419-6142-6_7. [25] C. Sommer, R. German and F. Dressler, "Bidirectionally Coupled Network and Road Traffic Simulation for Improved IVC Analysis", IEEE Transactions on Mobile Computing, vol. 10, no. 1, pp. 3-15, Jan. 2011, doi: 10.1109/TMC.2010.133. [26] Amoozadeh, M., “Vehicular Network Open Simulator (VENTOS)”, http://maniam.github.io/VENTOS. [27] Mani Amoozadeh, Bryan Ching, Chen-Nee Chuah, Dipak Ghosal, H. Michael Zhang, “VENTOS: Vehicular Network Open Simulator with Hardware-in-the-Loop Support", Procedia Computer Science, vol. 151, 2019, pp. 61-68, https://doi.org/10.1016/j.procs.2019.04.012. [28] M. Segata, S. Joerer, B. Bloessl, C. Sommer, F. Dressler and R. L. Cigno, "Plexe: A platooning extension for Veins", 2014 IEEE Vehicular Networking Conference (VNC), pp. 53-60, 2014, doi: 10.1109/VNC.2014.7013309. [29] Carl Bergenhem, Erik Hedin, Daniel Skarin, "Vehicle-to-Vehicle Communication for a Platooning System", Procedia - Social and Behavioral Sciences, vol. 48, pp. 1222 – 1233, 2012. [30] J. Posner, L. Tseng, M. Aloqaily and Y. Jararweh, "Federated Learning in Vehicular Networks: Opportunities and Solutions", IEEE Network, vol. 35, no. 2, pp. 152-159, 2021, doi: 10.1109/MNET.011.2000430. [31] Tengchan Zeng, Mingzhe Chen, Mehdi Bennis, "Federated Learning on the Road: Autonomous Controller Design for Connected and Autonomous Vehicles", arXiv:2102.03401, 5 Feb 2021. [32] Iftikhar Ahmad, Rafidah Md Noor, Ismail Ahmedy, Syed Adeel Ali Shah, Ibrar Yaqoob, Ejaz Ahmed, Muhammad Imran, "VANET–LTE based heterogeneous vehicular clustering for driving assistance and route planning applications", Computer Networks, vol. 145, pp. 128-140, 2018, https://doi.org/10.1016/j.comnet.2018.08.018. 280