Autonomous Vehicles - Science topic

Dear António Brandão,

I suggest you see the links on the topoic.

Best regards

Dear Keshav Koolwal,

I suggest you see the links on the topic.

Best regards

combining MARL and safety would be an interesting area

Optisystem from Optiwave Company

Technology Integration via;

1. Joint Sensing and Communication,

2. Dual Function Radar Communication.

Can you explain more about the topic, please?

Dear @Qamar Ul Islam, thanks for your response, however it is not the correct. Since the video explained the polar spiral or Archimedes spiral.

1. How about the role of 4D MMWR in robots and autonomous vehicles, especially compared with lidar and cameras? or even replace them as primary sensors?

It can substitue lidas or camera depending on the application.

2. Is 4D MMWR robust enough for L3 or above-level autonomous driving?

For Level 3 and up I would always use redundant sensors, preferentially radar, camera and lidar.

3. Are there any other fields that 4D MMWR will be applied more often or promising?

Several applications. There is no general answer.

For high-fidelity vehicle dynamics simulation, the gold standard should always be CarSim or ASM. For virtual environment creation, Unity may dominant.

The key difference between journal and conference paper is that a journal article is a long piece of writing that is published in journals and periodicals, whereas a conference paper is a concise and precise written paper that is presented in research conferences.

Both journals and conference papers present the findings and conclusion of a research study. Moreover, both journal articles and conference papers undergo a review process before the acceptance of papers.

Next, borrowing from Ian Kennedy: The conference paper gives faster feedback. The journal paper counts for about twice as much credit, but takes longer to get into press (an average of 9 months).

I recommend you contact the first author of the paper. You can find the email in the following page: https://www.nature.com/articles/s41586-018-0095-1

Kindly see also the following link: https://www.google.com/amp/s/www.theverge.com/platform/amp/2021/7/6/22565448/waymo-simulation-city-autonomous-vehicle-testing-virtual

Thanks for sharing.

Dear Shahid Ayub,

I suggest you to see links and attached files on topic.

Best regards

Hi Eric Ntiamoah . I agree with Qamar Ul Islam . Thanks.

Great question - given the attitude afforded by the Investigatory Powers Act in the UK, there is a huge privacy concern posed by CAVs. These are essentially mobile surveillance systems and the concern is how footage from the onboard interior and exterior facing cameras is used and whether state bodies can request access to these. Lots of really questions surrounding CAV use and development in the UK.

Very best wishes, James.

Dear Sicheng Wang

Good to Know, Good Luck.

Kind Regards

You can use SPSS or STATA or even statgraphics to code the data, analyze the data using multiple regression

Dear Abhijit Khadatkar

You are welcome. As I had already underlined, I have no knowledge about such cameras. I have never seen such cameras in my hand or any other place. I have not also seen any technical online video in detail about them . I know no detail about the cameras. I only suggested my idea through the theoretical scope. I discriminated between 3D cameras and 2D cameras. I had seen a video from Boston Dynamics Atlas robot and also in TV, in which some cameras render the environment image in 3D scan and can analyze the environment for collision avoidance. A 3D scanner camera is suitable for advanced motion planning. But I guess some simpler cameras you could possibly use for your case. I mean a camera which can distinguish between the rough terrain where the farms are bordered, and the traction where the crops are grown. This could be simply available by a camera which could detect two different colors: color of row crop, and the curb obstacles along borders of any farm. There might be no need for an advanced 3D scanner camera, so long as a simple camera can distinguish dark and light fields and tractions over the farming land based on their color (grey and black). This is all I knew.

Dear James Marson,

Thanks for the positive assessment of this discussion on the development of autonomous motor vehicle technology in the potential synergy of development with electromobility. Thanks for your comment. I agree with your view on this topic.

Regards,

Dariusz Prokopowicz

Dear Cao Wlight,

I suggest you to see links and attached files on topic.

Best regards

Did you mean to type pre-test?

The problem under discussion is obstacle avoidance by reinforcement learning. Reinforcement learning works based on the concept of reward based action. So, every time a obstacle is detected, your logic should be such that advancement or activation happens on a particular front which could be treated as the reward. Hope this helps a bit.

Yes I agree, there are bad humans who drive cars, design algorithms and write code, along with many other things. I assume you are not suggesting that robot makers are a cut above the rest of humanity and can make these determinations and dispense these permissions for the purpose of protecting the rest of us from bad drivers because we all hate (no argument there) bad drivers. How about killer robots and the authors of such protocols; call to:{contingent factors when it is necessary to choose whose death}

Is this an omelette making egg breaking argument because if so I find they tend to end by generalising about the quality of human types and their utility to other human types. Not a good outcome from my reading of history but who knows what the future may bring. Maybe we could build a robot better than human and all mass suicide for being slower dimmer lesser creatures. Who’d argue with that?

I would call it a network of autonomous rover. Their instruments will perform with a

better resolution and will be more versatile.

Will include deeper drilling, moving radar, etc. Also they will communicate with each other.

@tariq islam .Use dead reckoning.it is the process of calculating current position of some moving object by using a previously determined position, or fix, and then incorporating estimations of speed, heading direction, and course over elapsed time.

Farzad Torabi ,

I think both references target different problems altogether and it is not a matter of choosing doctrines. [1] is aimed at the problem that " the thesis that killing is worse than letting die cannot be used in any simple, mechanical way in order to yield conclusions about abortion, euthanasia, and the distribution of scarce medical resources." meanwhile [2] is aimed at autonomous vehicles.

Even putting this difference aside [2] states that "It is important to focus not on abstract thought experiments but on concrete conditions that influence the behavior and properties of self-driving cars as being developed through

different stages lading to deployment and inclusion in traffic". This also separates the scope of work of [2] with respect to [1].

I think [2] does not address the Trolley Problem and merely evades it. The statement "The relevant trolley problem scenario in the context of the state

of the art technology is thus more likely to be: hitting an obstacle that is correctly identified versus hitting another object that is unknown or incorrectly identified. It is unrealistic to assume that the self-driving car will have information about whether a human dies or not in a critical situation" is a different problem altogether. There are instances of the problem that take into account this issues and still the Trolley problem appears. further the stament in [2] that "When neural networks are trained by analyzing photos of pedestrians, are those photos subject to fairness?" is completely different than the Trolley problem.

In my sincere opinion I think that [2] should rename the title of the publication since it is a different set of problems that do not address the trolley problem, but are very important to take into account for AV design.

[1]Thomson, J. J. (1976). Killing, letting die, and the trolley problem. The Monist, 59(2), 204-217.

[2] Holstein, T., & Dodig-Crnkovic, G. (2018, May). Avoiding the intrinsic unfairness of the trolley problem. In Proceedings of the International Workshop on Software Fairness (pp. 32-37).

Please review and answer this question about Trolley problem in AVs

@Farzad possibly this could help your cause.

Yeshodara, N. S., Nagojappa, N. S., & Kishore, N. (2014, October). Cloud based self driving cars. In 2014 IEEE International Conference on Cloud Computing in Emerging Markets (CCEM) (pp. 1-7). IEEE.

U can use a Q function to learn values of making and not making a lane change and choose greedily. But for contious action space, Deep Q Learning is not preferred (but could be done, u need to predect mean and standard deviation and then sample), instead architecture like PPO is better for continious control

Hi,

This is a very interesting research domain.

Here are some sources:

J. Kong, M. Pfeiffer, G. Schildbach and F. Borrelli, "Kinematic and dynamic vehicle models for autonomous driving control design," 2015 IEEE Intelligent Vehicles Symposium (IV), Seoul, Korea (South), 2015, pp. 1094-1099, doi: 10.1109/IVS.2015.7225830.

Good luck with your studies.

Dear Martin Schmidt many thanks for the link - very helpful for our work.

Very best wishes, James.

Hello All,

I am searching for topics for master's thesis in the areas of ADAS, IoT.

Could someone please suggest my cutting-edge topics?

in case of any queries please mail me on qutibahhussein862@gmail.com

Thank you in advance

Best Regards,

Qutibah Hussein

Hi - in terms of the changes in the living environment i think the opening of access to convenient and perhaps personal means’ of transport for people with disabilities will be hugely significant in the introduction of CAVs. Also, given the move to electric vehicles, which most CAVs are likely to be, the benefits to clean air initiatives will also be important in peoples lives and their exposure to toxins in highly traffic-dense areas.

Hope this helps and very best wishes, James.

you can take a video stream on a straight line and then use motion vectors

Dear Reema Ahmad,

Thank you for your response. I agree with you. You added some very interesting information to our discussion. I also believe that the importance of artificial intelligence is growing in the implementation of this technology in various applications, in various fields, branches and sectors of the economy and in the improvement of research works. Thank you very much for proposing an article on this important issue artificial intelligence.

Greetings, Have a nice day,

Dariusz Prokopowicz

Hello Anum.

There are a plenty of agent-based traffic simulators. I used SUMO (https://www.eclipse.org/sumo) a couple of years ago, which is a powerful microscopic traffic simulator. I also recommend you this book of J. Barceló (https://www.springer.com/gp/book/9781441961419). It presents the basic of traffic simulation and gives details of many simulators.

On the other hand, there are some general purpose agent-based simulators. Many people use them to model and simulate traffic scenarios. Examples include NetLogo (https://ccl.northwestern.edu/netlogo) and Repast (https://repast.github.io).

Cheers,

Marcelo

Here our Towards Data Science article:

https://towardsdatascience.com/do-you-want-to-train-a-simplified-self-driving-car-with-reinforcement-learning-be1263622e9e?gi=606f63188da0

you can validate your data by using Dugoff's tire model which you made in matlab. in other world you should use adams matlab co-simulation. first made tire model in matlab. Then use this model as adams input. then run adams model.(your adams model with matlab tire model.)finally compare the adams vehicle model with matlab vehicle model.

Furthermore, you can use magic formula pacajka model to compare .It is more accurate.

Hi,

Also i suggest you to see links and attached file on topic.

Best regards

is there any reference about replacement of source after two half lives.

Dear Samira Eshghi,

The performance of the controller is generally measured using the following parameters such as :

1) Maximum Overshoot

2) Settling Time

3) Rise Time

4) Steady State Error.

Additionally, there are other ways to measure performance in literature. This includes integrated absolute error (IAE), the integral of squared-error

(ISE), or the integrated of time-weighted-squared-error (ITSE). These performance measures have their own advantages and disadvantages. For example, the disadvantage of the IAE and ISE criteria is that its minimization can result in a response with relatively small overshoot but a long settling time because the ISE performance criterion weights all errors equally independent of time. Although the ITSE performance criterion can overcome the disadvantage of the ISE criterion, the derivation processes of the analytical formula are complex and time-consuming. The formulas for these performance measures or the performance measure based on overshoot, settling time, rise time and steady-state error can be found in the paper given below.

In discrete domain, ISE refers to Mean Square Error. I am unfamiliar with normalized-mean-square-error .

Nice Dear John Mendy

Thank you very much Sai R N.I am most grateful.

It seems that the paucity of innovations in the field of electric motors, especially in recent years, has become a big concern for the researchers. To the best of my knowledge, this concern has become rampant to a great extent, however, I think vice versa. New types of electric motors are developed by a lot of researchers for a broad variety of industrial applications with a focus on electric vehicles, including VFRMs, VRMs, FSPMs, Memory machines, and etc, all of which are proposed to increase efficiency, power and torque densities, and decrease the consumption of rare-earth PMs. Innovation Continues. The following paper is a good source to support this opinion.

Both have pros and cons:

https://youmatter.world/en/hydrogen-electric-cars-sustainability-28156/

The wheel torque can be interpreted as throttle opening percentage by modeling the driveline backwards, i.e. wheels=> driveshaft=> gearbox=> move-off element (clutch or torque converter)=> engine, and finally the engine map can be used to obtain the throttle opening percentage. A similar approach for brake system can be explored as well.

I know it seems obvious, but you should perform this procedure using the parameters defined in your validation platform (IPG). Otherwise, you should find the detailed specifications of your considered vehicle.

I suggest to start literature review related to your objectives.

Read the last 5 to 10 years most related published research works closer to your topic. In particular carefully read Conclusion and Future work section of each of these (5 to 10 years published work) papers, where new research directions proposals or open problems will be available for you.

Best of luck for your literature review!!!

Ankit R. Patel I think the developers also sometimes underestimate peoples wishes and will to own and to control a vehicle. There is a deep feeling of freedom in driving a vehicle by your selves. The autonomous vehicle will probably be safer for the driver and the surroundings, but that has never stopped humans from wanting their own control.

Seems like a good idea. If the operation of the AVs is hugely successful, they could be integrated with the general vehicles. If it's not successful at all, the lanes could be converted to general vehicle lanes at little cost. In the process we will all increase our knowledge of the risk associated with the operation of AVs.

Dear Victor Egold, the reason why RTK is more accurate than DGPS is because of the method of the range measurement. RTK uses carrier phase-shift measurement while DGPS uses code measurement. In carrier phase measurement, the integer ambiguity is resolved. That is, the fractional part of the cycle in the signal is resolved. This resolved fractional part of the cycle is added to the total number of complete cycles before multiplying it with the wavelength to obtain accurate/precise range from the receiver to the satellite. The ambiguity is determined using three techniques: Single Differencing, Double Differencing and Triple Differencing. All of these techniques require that additional observations be obtained.

Fully autonomous driving is still an area of research. However successful implementation of various ADAS modules and and invent of advanced communication and signal processing technologies used is Connected Vehicles can make it a reality in near future. To relate ADAS features to automation level, the U.S. department of transportation (DOT) has recently released a guideline for safe testing and deployment of automated vehicles. DOT gives a “Levels of Automation” classification for driverless cars as a guideline in SAE international [1].

The embedded processors within the vehicles are interacting constantly with environment and gathering enormous amount of data which in tern may lead to a centralized level supervision for the vehicles.

[1] SAE On-Road Automated Vehicle Standards Committee. "Taxonomy and definitions for terms related to on-road motor vehicle automated driving systems." SAE International (2014).

Dear Samira Eshghi,

I suggest you to see links and attached files on topic.

Feedback Control of a Nonholonomic Car-like Robot - LAAS

https://homepages.laas.fr/jpl/promotion/chap4.pdf

Exponential stabilization of mobile robots with nonholonomic constraints

https://ieeexplore.ieee.org/iel4/9/4443/00173153.pdf

Control of a Nonholonomic Mobile Robot: Backstepping ... - CiteSeerX

citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.436...

Path planning and control of non-holonomic mobile robots

https://www.theses.fr/2013ECLI0025.pdf

Best regards

In any case (including High Tech / AI supported cars), I believe that the minimum age to allow getting a driving license should be 18 years old.

I think it will take long time (may be more than 10-20 years), as there are number of factors affecting/not clear yet, it includes:

1. Government policies

2. Acceptance and Trust level by the common people

3. Rules for the autonomy, specifically who will responsible for any accident, if any

4. Cyber Security threats

5. Road conditions

Thank you.

Ankit Patel

The autonomous driver cars can not completely substitute the manually driven cars. This may be an ultimate goal. Hybrid driven cars may be more realistic goal for the next time.

Best wishes

I conclude that the idea that one can transfer control of the almost autonomous vehicle to the human occupant when the going gets too complicated for the system is fatally flawed. Driving skill assumptions about the hapless occupant are bound to be inaccurate and far lower than a well designed electronic system. If the situation is starkly perilous, the "operator" may well have fainted dead away. The closer the development process gets to the completely autonomous (level 5) design, the more insurmountable the handover. Even a moderately well designed level 5 system will be able to outperform 99% of the anticipated human operators. The user experience in the autonomous vehicle will, of necessity, be that of a train ride with no user concern for the technical or practical matters of the vehicle. The user experience in the almost autonomous (level 4.8 or so) vehicle may be so disconcerting as to prevent anyone who is informed of the control transfer situation from acquiring such a vehicle, or taking a ride taxi style.

John M. Kirkwood, CHFP

2208 Hunter Street Cinnaminson, NJ 08077

Phone: 609-828-2045

email: john.kirkwood@mac.com

Eduardo Corrêa Gonçalves

Thanks very much

The paper by Yann LeCun pointed to by Eduardo sufficiently answers my question. It is now clear that BP has not been replaced and remains important down to practical real world applications. From LeCun paper and after some navigation I found the following additional references:

Ian Goodfellow, I. et al Deep Learning (book by MIT Press)

Jürgen Schmidhuber, Deep Learning in Neural Networks. An Overview

Cireçan, D. et al - Flexible, High Performance Convolutional Neural Networks for Image Classification

Time ago I published the paper Generalized Backpropagation containing a deduction of general BP formulas

and with BP still useful may now write a further generalization.

Hearty thanks to all, especially to Eduardo.

Daniel Crespin

I am doing research, maybe i can become the part of company for research direction and collaborate in research field, if you are interested, we can talk more through this channel, my email id

Thank you Mohamed-Mourad Lafifi

Friction circle (based on the assumption that a tire can develop equal limiting longitudinal and lateral forces) gives you:

- The maximum longitudinal force that the tire can develop while it is developing a specified lateral force, or

- The minimum lateral force that the tire can developed while it is developing a specified longitudinal force.

Consequently, the maximum longitudinal force in the absence of lateral force generation and the maximum lateral force in the absence of longitudinal force will both be equal to adhesion coefficient*normal tire load.

For tires developed for emphasized traction or cornering capability, "friction ellipse" may be more suitable.

Dear Arash,

Please take a look at links and attached files in topic.

-Vehicle Dynamics of Modern Passenger Cars

-Camber Angle Control Method Corresponding to the Electric Vehicle ...

-Camber-Car Design & Dynamics - Zonagravedad

https://www.zonagravedad.com/.../Carrilanas/Camber_Car_Design...

-)) (( tan)/() )( 1( )) ( tan sin( S B BE S E S B C D F + + + −=Φ +Φ = α α γ ...

-performance et comportement des vehicules - ULg - LTAS

Best regards

Dear Valentas,

It should be pointed out that the data structure technique belongs to the large domain of mathematical optimization, and combinatorial optimization is an element of this set. In addition, heuristics and metaheuristic techniques are part of this domain and are an extension for the resolution of multiobjective optimization. It is clear that these techniques can be used in a complementary manner or in hybridization.

For more information about this sbject, i suggest you to see links and attached file in topic.

How Autonomous Vehicles are Driving a Shift in Supply Chains

https://medium.com/self-driving-cars/how-autonomous-vehicles-ar...

Solving Combinatorial Optimization Tasks by Reinforcement Learning ...

www.cs.ubbcluj.ro/.../Optimization.../solving_combinatorial_optim...

Logistics 2050 A Scenario Study - DHL

big data in logistics - DHL

www.dhl.com/content/dam/.../g0/.../CSI_Studie_BIG_DATA.pdf

Best regards

Dear see these topics:

Thank you very much for the links and files attached professor Lafifi.

I suggest you check the affordable Texas Instruments Evaluation Module AWR1642BOOST which incorporates the 77- and 79-GHz FMCW TI-AWR1642, or his brother EVM system based on the TI-IWR1443 which provides elevation besides azimuth (bearing) data. Texas has some nice demos of these systems available online and, besides the software develop tools, provides browser-based GUIs for testing the system.

To your point and as I think of autonomous or remote-controlled, the differentiation in my mind is having a human driver-in-the-loop for autonomous vehicles. Remotely controlled has access for “in situ” data and an observer to command maneuvers. However, the controls and actuation for an on-ramp entry maneuver into congested freeway, for example, is quite challenging and requires a driver-in-the loop for safe driving operations. As the environment or weather conditions could be classified as degraded, the aforementioned research problem continues to exist. I agree that range and range-rate information for targets in the field-of-view are reported are reported by RADAR; however please expand on sensors necessary for capturing and reporting image data and to provide insights about a degraded visual environment.