Tag: #AutonomousCars

Autonomous Driving Using AI

Posted on by Blogs

Premium Vector | Car modern interior, cockpit view inside. illustration. artificial intelligence.

Tech companies and the auto industry are working hard in tandem to make autonomous driving a reality by the early 2020s. Driverless cars with various levels of human participation will roll out in stages over the next few years, with fully-autonomous SAE Level 5 driving on the scene by 2030.

Today, most automotive manufacturers have achieved Level 2 assisted driving where the car can manage simple scenarios, like active lane centering and parking assistance, itself. Fewer manufacturers provide Level 3 autonomous driving where the car can autonomously navigate a traffic jam or roadways to a destination. For both levels, human drivers can take the wheel if they choose.

Everything to know about the future of self-driving cars - Maclean's

The limitations of AI that prevent advancement to fully-autonomous driving

From an engineering standpoint, Level 3 autonomous driving is powered by two things: hard-coded structured programming models mostly written for embedded systems and deterministic rules that make decisions supported by neural networks.

These two things combine to build AI driving agents, but with at least five important limitations:

  1. Lack of perception and behavior intelligence compared to humans. Unlike existing AI agents trained with machine learning (ML), humans don’t need thousands of images of trees, for example, to recognize a tree or identify a driving situation.
  2. Low accuracy performance. With existing tools, the probability of steering accuracy decreases as more autonomous driving functions and components get added. The complex real driving system will deliver only about 60 to 70 percent accuracy performance on motion control for steering and acceleration, well short of what’s required for fully autonomous driving.
  3. Inability to cope well with complexity. Deterministic rules are usable in closed environments, such as a contained driving course, but can’t capture the complexity of real-world driving situations.
  4. Require too much data. Usually ML models require enormous amounts of data, which are too expensive and difficult to collect and move over existing corporate networks.
  5. Take up too much run-time, CPU/GPU processing and address exabytes of storage resources. It takes a lot of time and power to process the large volumes of automotive data and learn from them – and that’s often not cost-effective.

For the industry to evolve toward fully-autonomous driving, technologists have to develop an AI model that mirrors human driving behavior. In doing so, we need to guarantee a deterministic behavior that always produces the same result from the same input. The industry needs a new approach.

The big question then remains: How will the car act autonomously and intelligently in real time, in the real world?

We believe a big part of the answer lies in adapting knowledge of the human brain to AI – and we have drawn much of the inspiration for our new approach from brain science research conducted by Danko Nikolic at the Frankfurt Institute for Advanced Studies (FIAS) in Germany .

Adapting innovative brain research to AI and the production of fully autonomous vehicles has emerged as one of the more exciting technology innovation breakthroughs we’ve seen in the past couple of years.

While it will take time, the benefits to society of producing self-driving vehicles – and simply learning more about how the brain works along the way – holds great promise for human progress.

Still No viable method of transporting data from autonomous cars tests

Posted on by Blogs
Driverless Cars

Behind the scenes at locations around the world the auto makers are running tests on autonomous cars for literally thousands of hours. The industry has poured more than $80 billion into R&D on autonomous cars over the last four years, so they are serious about making this happen.

Those of us working on these tests have one overwhelming challenge: how to manage all the data that gets generated during the tests. One eight-hour shift can create more than 100 terabytes of data. In a week of testing multiple cars, we’re talking about petabytes of data. And often — at rural testing centers, for example — Internet bandwidth speeds are simply insufficient to ensure that the data reaches our data centers in North America, Europe and Asia at the end of the test day.

Autonomous car

Right now, we have two main ways to transport data back to a data center. They are both cumbersome, but have different plusses and minuses. Until advances in technology make these challenges easier to manage, here’s what we do today:

  • Connect the car to the data center. Test cars generate about 28 terabytes of data in an hour and it takes 30 to 60 minutes to offload that data by sending it to the data center over a fiber optic connection. While this is a time-consuming option, it remains viable in cases where the data gets processed in somewhat smaller increments.
  • Take/ship the media to a special station. In many situations the data loads are too large and the fiber connections unavailable (e.g., at geographically remote test locations such as deserts, ice lakes and rural areas) to upload data directly from the car to the data center. In these cases we remove a plug-in-disk from the car and take it or ship it to a “Smart Ingest Station” where the data is uploaded to a central data lake. Because it only takes a couple of minutes to swap out the disks, the car stays available for testing. The downside of this option is we need to have several sets of disks, so compared to Option 1 we are buying time by spending money.

In three to five years we may get to the point where both options are outmoded by advances in technology that make it possible for the computers in the car to run analysis and select the needed data. If the test car could isolate the test-car video on, for example, right-hand turns at a stop light, the need to send terabytes of data back to the main data center would be alleviated and the testers could send these smaller data sets over the Internet.

Of course, we’re several years away from having such a capability. In the past year, IBM and Sony have been working on a 330 terabyte tape drive that promises faster and more resilient data storage in a form factor that can fit in the palm of your hand. Once such products are commercialized, it should make our lives a bit easier.

Ultimately, we’d like the ability to move our various equipment easily in and out of hotel rooms and carry it on plane trips in our pockets or briefcases. Today, the equipment is often clunky and hard to move around. While technology can help, we have to be realistic and understand the data challenges surrounding autonomous cars are likely to increase exponentially.  The challenges may grow, but at least sometime soon the gear we use won’t be so cumbersome that our muscles ache at the end of the day.

Ways to better data processing in Self-Driving Cars

Posted on by Blogs

autonomous vehicle development

Autonomous cars promise to change the face of transportation, offering many more mobility options for individual motorists and companies alike. In moving forward with this new technology, our automotive clients have a very important challenge to overcome: processing the petabytes of data that gets collected during the development and testing of autonomous driving systems.

KPIs have always been important to car makers. They are necessary to attain road approvals and to track key competitive differentiators. With autonomous cars, however, car makers are accumulating – and must find ways to process and manage – 10, 20, sometimes 30 times the data as before.

As a result, they need much more efficient data analysis tools that can help them analyze the data for the specific autonomous car KPIs they are looking for. To make this happen, they need to take the following four steps:

  1. Make sure the car’s sensors are working. There are typically eight to 12 sensor systems in an autonomous vehicle test car. It’s important to look at the data at the very beginning of the workflow by checking the KPIs to ensure that the system works properly. Some of the KPIs car testers evaluate include the following: vehicle operations, safety, environmental impact and in-car network efficiency.
  2. Scale the workflow to process the data. Traditional architectures of automotive frameworks are not suited for the large-scale data processing workloads required for testing the algorithms used in autonomous car tests. In using traditional data storage methods, vehicle test data gets stored in NAS-based storage and gets then transferred to workstations, where engineers test algorithms under development. This process has two downsides:
    • Large amounts of data must be moved, requiring considerable time and network bandwidth.
    • Individual workstations do not offer the massive computing power required to return test results fast enough.

    Today, testers are extracting each frame of video data with its associated Radar, Lidar and sensor data by using open source Hadoop. The major benefit of Hadoop is that it scales processing and storage to hundreds of petabytes. This makes it a perfect environment for testing autonomous driving systems.

  3. Make the most of data analytics. In processing petabytes of automotive data, we have to look at how we present the data to higher level services. New data analysis tools can read different automotive formats to give us proper levels of access to the metadata and data. For example, say we have 700 video recordings, we now have tools that can pinpoint footage from the front-right camera alone to show how the car performed making right-hand turns. We can also use the footage to determine the accuracy of a model depicting the autonomous car’s perception of its ambient physical surroundings .
  4. Run the data analysis. In the end, we want to use data analysis tools to give R&D engineers a complete view of how the car has performed in the field. We want to generate information on how the systems will react under normal driving conditions.

Overcoming these data analysis challenges is critical. Manufacturers can’t obtain permits for releasing their cars until they can show that the cars performed up to certain standards in road tests. And when autonomous cars do start to hit the roadways in the next few years, auto manufacturers might need the KPIs they generated in testing. A few accidents are inevitable and, when questions arise, car makers can use KPIs to show the authorities, insurance companies and the general public how the cars were tested and that proper due diligence was performed.

Right now, there’s some distrust among the driving public of autonomous cars. It will take a massive public relations effort to convince consumers that autonomous cars are safer than traditional manually-driven cars. But proving that case all starts with the ability to process the data more efficiently

Autonomous cars promise to change the face of transportation, offering many more mobility options for individual motorists and companies alike. In moving forward with this new technology, our automotive clients have a very important challenge to overcome: processing the petabytes of data that gets collected during the development and testing of autonomous driving systems.

KPIs have always been important to car makers. They are necessary to attain road approvals and to track key competitive differentiators. With autonomous cars, however, car makers are accumulating – and must find ways to process and manage – 10, 20, sometimes 30 times the data as before.

self-driving vehicle technology

As a result, they need much more efficient data analysis tools that can help them analyze the data for the specific autonomous car KPIs they are looking for. To make this happen, they need to take the following four steps:

  1. Make sure the car’s sensors are working. There are typically eight to 12 sensor systems in an autonomous vehicle test car. It’s important to look at the data at the very beginning of the workflow by checking the KPIs to ensure that the system works properly. Some of the KPIs car testers evaluate include the following: vehicle operations, safety, environmental impact and in-car network efficiency.
  2. Scale the workflow to process the data. Traditional architectures of automotive frameworks are not suited for the large-scale data processing workloads required for testing the algorithms used in autonomous car tests. In using traditional data storage methods, vehicle test data gets stored in NAS-based storage and gets then transferred to workstations, where engineers test algorithms under development. This process has two downsides:
    • Large amounts of data must be moved, requiring considerable time and network bandwidth.
    • Individual workstations do not offer the massive computing power required to return test results fast enough.

    Today, testers are extracting each frame of video data with its associated Radar, Lidar and sensor data by using open source Hadoop. The major benefit of Hadoop is that it scales processing and storage to hundreds of petabytes. This makes it a perfect environment for testing autonomous driving systems.

  3. Make the most of data analytics. In processing petabytes of automotive data, we have to look at how we present the data to higher level services. New data analysis tools can read different automotive formats to give us proper levels of access to the metadata and data. For example, say we have 700 video recordings, we now have tools that can pinpoint footage from the front-right camera alone to show how the car performed making right-hand turns. We can also use the footage to determine the accuracy of a model depicting the autonomous car’s perception of its ambient physical surroundings .
  4. Run the data analysis. In the end, we want to use data analysis tools to give R&D engineers a complete view of how the car has performed in the field. We want to generate information on how the systems will react under normal driving conditions.

Overcoming these data analysis challenges is critical. Manufacturers can’t obtain permits for releasing their cars until they can show that the cars performed up to certain standards in road tests. And when autonomous cars do start to hit the roadways in the next few years, auto manufacturers might need the KPIs they generated in testing. A few accidents are inevitable and, when questions arise, car makers can use KPIs to show the authorities, insurance companies and the general public how the cars were tested and that proper due diligence was performed.

Right now, there’s some distrust among the driving public of autonomous cars. It will take a massive public relations effort to convince consumers that autonomous cars are safer than traditional manually-driven cars. But proving that case all starts with the ability to process the data more efficiently

Delivering excellence, collaborating across time zones.

Take a look at our global hideouts.​

DMCA.com Protection Status

Contact

contact@anteelo.com

Twitter Instagram Dribbble Linkedin Facebook

India (HQ)

C-2073A, Sushant Lok, Phase-1, Gurgaon,
Haryana 122002

Atlanta, USA

120 West Trinity Place Decatur, GA 30030

London, UK

33 St James’s Square, London
SW1Y 4JS

Dubai, UAE

704, Saheel Tower 1 Al Nahda 1,
Dubai, UAE

Melbourne, Australia

291 -Blackburn Road, Burwood East Vic 3151

Surabaya, Indonesia

Jl. Hang Tuah 14 Sidoarjo Jawa
Timur 61212

India (HQ)

C-2073A, Sushant Lok, Phase-1,
Gurgaon, Haryana 122002

Atlanta

Atlanta, USA – 120 West Trinity Place
Decatur, GA 30030

London

33 St James’s Square,
London SW1Y 4JS

Dubai

UAE – Office # 704, Saheel Tower
1 Al Nahda 1, Dubai, UAE

Australia

Level 33, Australia Square 264
George Street, Sydney, 2000

Indonesia

Jl. Hang Tuah 14 Sidoarjo
Jawa Timur 61212

© 2018-2022 Anteelo Design Private Limited
A Quantum Oakmen Partner Venture

error: Content is protected !!