Two Second Advantage: What if you could look into the future?

BlueSpace was founded with the thesis that improving the reliability and accuracy of perception will unlock autonomy and unleash commercialization, adoption, and trust in the technology.

In the previous blog we discussed how BlueSpace's industry-leading motion accuracy software outperforms the industry and is critical for safe and scalable ADAS/AD applications.

While current approaches infer object motion based on position changes and heading angles over multiple frames, our software accurately measures the full motion state of all objects in a single frame of data, without needing position or heading angles.

BlueSpace discerns the motion signals of objects that we could not pick up previously with machine learning approaches to perception and prediction. BlueSpace’s 4D Predictive Perception derives these most subtle, yet important movements as early as two seconds before other methods. Compared to the typical perception stack, which first tries to understand what an object is, we focus on first observing the consistent motion and deriving the objects and motion states that best explain the data. [more details here].  

Why is this important? Why should you care? Things happen - unexpected things - when driving in the real world. Here are examples of when knowing motion is more critical and essential than knowing what object it is to infer motion (i.e., do you really need to know that an Object X fell out of the truck in front of you vs. what-ever-it-is is in my path coming at me with Y motion)

Conventional approaches to AD rely on training. So “if” the computer has seen it before, it “may” know what to do. Well, what if the AV stack has not trained sufficiently enough to react to an unexpected event, especially in a timely way? The fact of the matter is that one could never train enough to catch up to dynamic reality. One can think of myriads of novel objects, combinations of objects, and unexpected events on the road that would be nearly impossible to comprehensively address with training: falling and occluded objects, erratic drivers/pedestrians, novel mobility types, etc. The long tail is indeed very long.

With BlueSpace’s highly precise motion parameters and generalizable approach to objects, ADAS/AD systems can detect ultra-fine movements for any potential road hazard - even if it’s never been seen before. For example, a yaw parameter that describes an object's rotation and is a crucial indicator to determine its path and whether an object will turn into the ego vehicle's path. A vehicle not equipped with BlueSpace's perception software typically uses an object’s position, which in many cases is too late. These limitations leave the motion planning system with very little reaction time. At best, the ADAS slams on the brake at the last minute, or it fails to activate early enough, thus not being able to avoid a collision.



Note: Plots displayed below are best viewed on a desktop browser


BlueSpace's Two-Second Advantage - Enabling Safety and Comfort for ADAS/AD

Here’s a simplified yet pointed example of how BlueSpace’s foresight avoids a collision by deciphering full-motion based on our software’s ability to discern subtle movements of the object. In the simulation below, we showcase how our technology equips the vehicle with a time advantage so that it can avoid a full-on collision, potentially fatal.

If autonomy were to promise a safer future - let’s show the world how it can be done. BlueSpace demonstrates our motion-first approach, allowing critical time advantage to take safe action in unprecedented and unexpected situations. These additional milliseconds secured can make all the difference between collision or non-collision, or life and death. BlueSpace's time advantage is a game-changer to enhance safety ADAS/AD.

BlueSpace's perception software detects the precise motion early, providing time advantage and also predicts the object's trajectory so that the ego vehicle can take safe action. Precise, low latency measurement of full-motion equips ADAS/AV with an advantage to make timely safety-critical decisions. These milliseconds can make all the difference between life and death. BlueSpace's time advantage is a game-changer to enhance safety ADAS/AV. 


Experience BlueSpace's 4D Predictive Perception for yourself. Reserve your spot by - emailing partners@bluespace.ai

Embed Block
Add an embed URL or code. Learn more
Previous
Previous

Why Diversity Matters in Autonomy

Next
Next

Billion Dollar Question: How good is the perception system in your AV?