Cart 0
Cart 0
 
RiskEngine-02.png
 

A safety compliance tool to help you generate confidence in autonomous safety. 

Retrospect’s RiskEngine™ is an independently developed and universally available safety metric that provides a cross-industry measure for safe driving behavior in self-driving cars. RiskEngine™ can be used for safety metrics, such as “Safety Performance Indicators” or SPIs, and for speeding up internal and external safety approval processes.

What will RiskEngine help you accomplish? 

Start tracking risk

start_tracking_risk_graphic-01.png

Don’t wait for the dust to settle on final safety metrics or SPIs before you start tracking AV risk. RiskEngine™ helps you leverage past and current driving data to meet the safety standards of tomorrow. Let us automatically update the safety metrics for seamless backwards compatibility, allowing you to get the most from each byte of data you generate and make the safest self-driving cars out there.

 

CLASSIFY SCENARIOS & Behavior

scenario_agnostic_risk_graphic-01.png

RiskEngine™ is scenario agnostic, meaning it can be used across any scenario. Starting at the Ground Truth level, RiskEngine™ provides the base calculations for quantifying the Potential Risk in any given scene. This not only gives the industry a common language when it comes to evaluating risk, it provides an objective measure to compare scenarios and to characterize driving behavior.

 

FAST-TRACK SAFETY APPROVALS

share_results_risk_graphic-01.png

Speed up internal and external safety approval processes by using RiskEngine™ to create your safety reports. You can instill trust in your customers when they know your development and operations are independently reviewed with RiskEngine™ methods. RiskEngine™ metrics can be easily shared with road authorities and government agencies quickly, consistently, and without compromising IP.

 

 

 
“Today, neither industry nor government can assess the safety of self-driving cars.”
— EE Times, ‘A Wave of Safety Standards to Hit in 2020’
 
shutterstock_1341877205 copy.jpg
 

Try RiskEngine™

 
riskengine_steps-02.png

Step 1: Download ONE OF The sample scenarioS (or use your own)

We have provided two sample datasets courtesy of levelXdata, with permission.
You can also create your own existing datasets. Your data should follow the same format shown in our examples.

If you need assistance in importing your existing data, please contact us for support.

 

Example Dataset 1

Car makes a right turn

 
 
 

Example Dataset 2

Car turns right and collides with outside agent

 
 
Thanks to our friends at levelXdata for supplying example scenarios

Thanks to our friends at levelXdata for supplying example scenarios

riskengine_steps-03.png

Step 2: Upload Data (.csv)

Click on the Upload Data button, below. Upload your own data, or use one of our examples above.

If you need assistance importing your existing data, please contact support.

 
 
riskengine_steps-04.png

Step 3: check your email for a customized REPORT

Within a few minutes, the report will be sent to your inbox with the RiskEngine™ Potential Risk Data (csv) and Summary Report (pdf) attached.

If you’re interested in even more insights from the data provided, please let us know!

Also, we would love to hear from you on how we could improve RiskEngine™ even further.

riskengine_steps-05.png

STEP 4: SHARE YOUR RESULTS

We need your help in making autonomous safety measurements intuitive and widely understood by the autonomous community.

Tell others about the insights you’ve found with RiskEngine™ on LinkedIn. If you would like to incorporate RiskEngine™ into your toolchain and safety approval process, please let us know.

Let’s create a more ethical and transparent safety development process together!

 

Do you have suggestions or other resources you’d like us to consider? Contact us, below. We’d love to hear from you and improve the safety measurements for everyone.

 

Want to Learn More?

We’ve provided additional resources to help you extract the most value possible from RiskEngine™. 

You can also contact us and schedule a free consultation and we would be happy to discuss all of your questions.

→ RiskEngine™ Theory

What is Risk?

We use the industry and functional safety standard definitions for risk:

sample bandaged hand.png

Risk is the “combination of the probability of… physical injury or damage to the health of persons… [and] estimate of the extent of harm…”

(ISO 26262)

Currently, RiskEngine™ focuses exclusively on the risk of injury due to a collision. Collision hazards are the most common type of risk we think about when it comes to autonomous vehicles. Other hazard sources that will be included in later releases of RiskEngine™ are:

  • Motion discomfort and injury

  • Roll-over

  • Startle and fear

  • and many others…

We further define risk from collision into three helpful layers, according to the level of uncertainty being considered:

  1. Potential Risk

  2. Predicted Risk

  3. Planned Risk

What ARE THE RISK “LAYERS?”

Potential Risk is most closely related to the concept of kinetic energy, which is the primary source of the energy which causes harm in a collision. Potential Risk takes a control-neutral approach to evaluating risk. The goal of Potential Risk is to say, “All other events leading up to this moment have created the kinetic energy at this exact instance. What is the current energy at this given instance, where is it directed, and how might it contribute to injury in a collision?”

Predicted Risk builds another layer of risk on top of Potential Risk. Predicted Risk looks at kinematic models which are highly suitable for the current dynamic scene. The Predicted Risk may provide a guaranteed confidence level, e.g. 95%, over a given time horizon, or over motion boundaries, or both. The goal of Predicted Risk is to address the following question, “If I were watching a traffic scene and that scene were suddenly paused, how would I sketch out an animation of where the traffic participants will be over the next 3-5 seconds?”

Planned Risk is the top layer of risk that builds on Predicted Risk. Planned Risk incorporates a known trajectory into the predicted scenario. Predicted response models or context-based predictions (such as eye-contact or turn signal) may be included with convenient, event-based confidence intervals, such as, “E4 - High Probability” to “E1 - Very Low Probability.” The known trajectory is analyzed for achievability or “Controllability” of the actuation, using conveniently tiered confidence intervals, such as, “C1 - Simply Controllable” to “C3 - Difficult to Control.” The goal of Planned Risk is to answer the question, “How confident am I that I can achieve this trajectory based on my actuator limitations? And if I failed to follow the trajectory, what is the worst that could happen?”

 
shutterstock_580028608.png
 

Get Started with RiskEngine™ Today