Fwd: Act Now: Require Automated Vehicles to See Bicyclists and Pedestrians

[K Griswold]

I just signed this letter and encourage you to consider signing it.

Kathy Griswold

---------- Forwarded message ----------
From: League of American Bicyclists <communi...@bikeleague.org>
Date: Tue, Mar 20, 2018 at 2:09 PM
Subject: Act Now: Require Automated Vehicles to See Bicyclists and Pedestrians
To: grisw...@gmail.com

Tell your Senators to make automated vehicles pass a vision test. View this email in your browser Dear Bike Advocate: On Sunday night, an autonomous vehicle hit and killed a pedestrian in Tempe, AZ, as she was walking her bicycle across the street. While the details of the crash are still forthcoming, the League of American Bicyclists is concerned that these vehicles are being deployed without having to first prove their ability to recognize and respond to people biking and walking in our streets.  When human drivers apply for a driver’s license we have to pass a vision test. The League believes that all automated driving systems should first have to pass a “vision test” as well — requiring a safety performance standard — proving their ability to recognize and respond to people bicycling and walking, before they are on community streets. Right now the Senate is considering S. 1885: AV START Act to set guidelines for automated vehicle manufacturers to test their vehicles on our streets. Please join the League in asking Senators to require automated vehicles to pass a vision test. Send a message to your senators now! Share Tweet Forward Copyright © 2018 League of American Bicyclists, All rights reserved. Our mailing address is: 1612 K St NW, Ste 1102 Washington, DC 20006 Want to change how you receive these emails? You can update your preferences or unsubscribe from this list

--

Katherine J. Griswold

Michigan MBA & MSW

734.657.7900

[Ken Freeman]

Are we actually asking for an "eye test?"  What is needed is to require vehicle-level confirmation testing that the standards (ISO standards, at least) for forward vehicle and other forward object collision warning and collision mitigation responses shall be complied with, in tests which are also standards.  This compliance has to be established for Levels 1 through 5 autonomy.  If a vehicle is earlier equipped with (for example) an Automatic Emergency Braking (AEB) capability or a Forward Vehicle Collision Mitigation capability, then it has to be confirmed for Level 0 vehicles as well.

The same goes for more recent Pedestrian detection functions.  

In addition the action of a higher-level vehicle system to alert a safety driver or simply the human occupants must be developed according to the  Human Factors Engineering standards and best practices and confirmed to be effective in garnering an appropriate emergency control response.

The strategy in these standards is not to test for detection, because detection cannot be detected in a test situation.  But if warning or avoidance countermeasure action is correctly executed in an honest test, that can be observed, and it cannot (barring random coincidence) be achieved if the object detection and recognition functions are deficient.

The industry has already established that to test the performance of the such countermeasure systems is more effective and more feasible than testing for detection of external objects.  How the detection and object recognition are performed is the business of the design team that engineers the collision warning, mitigation, or prevention system.

An "eye test" which is fancifully named might be thought to be better than nothing, but it might actually be much worse than nothing.  It might also be all we get, which could be disasterous.

Ken Freeman

--
Please note that WBWC google groups is an unmoderated forum. It was developed by the Washtenaw Bicycling and Walking Coalition as a tool for bikers and walkers to discuss key issues and share information. However, not all views expressed in this group are the views or values of WBWC. For more information about WBWC, meetings, and projects, please visit www.wbwc.org
---
You received this message because you are subscribed to the Google Groups "Washtenaw Bicycling and Walking Coalition" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wbwc+unsubscribe@googlegroups.com.
To post to this group, send email to wb...@googlegroups.com.
Visit this group at https://groups.google.com/group/wbwc.
For more options, visit https://groups.google.com/d/optout.

--

Ken Freeman
Ann Arbor, MI USA

[Ken Freeman]

Where can the text of the S.1885 be found?

[John Lindenmayer]

https://www.congress.gov/bill/115th-congress/senate-bill/1885/text

John Lindenmayer
Executive Director

League of Michigan Bicyclists

410 S. Cedar St. Suite A
Lansing, MI 48912
(517) 334-9100
www.LMB.org
Twitter  |  Facebook

‌

[Ken Freeman]

Dear Mr. Lindenmayer,

I don't know where the LAB, LMB, or the US Senate now stand regarding this letter and S.1885, but I think the request of the cyclists and pedestrians should be different, something more technically feasible.  We ask for an "eye test" to be required.  I believe, based on my deep experience in the relevant automotive fields this touches, that this is a requirement that cannot be fulfilled.  It will require measuring information that is not publicly accessible in production vehicles, and will most likely be regarded as proprietary by the car companies and AV companies.

if AV (level 2 or higher) or more primitive autonomous systems (Levels 0 and 1) are tested, the results will not be meaningful if not as samplings of production cars, and for developmental vehicles to be tested as if they are production vehicles.  The established car companies (the OEMs) all understand this theoretically and practically.  The emerging car companies (Waymo, Uber, and others) which have chosen staff from the OEMs also understand this, because they brought in the correct expertise.  

To make the necessary performance tests on production vehicles, test data must be read from a common, open data network that all OEMs use and for which the testers can buy equipment - not a proprietary data network!  CAN is the most common example at this time in the industry.  Test technicians trained to follow NHTSA test requirements must be able to simply bring equipment and plug it in to the car.  Data that indicates response to a target by the vehicle system, such as triggering of a mild deceleration (adaptive cruise control), a driver warning (forward vehicle collision warning), or applying aggressive emergency braking (forward vehicle collision mitigation) is already standardized to be present on the CAN bus, but information that would characterize detection and its quality would usually not be present.  If the test scenario has only one object and it is a pedestrian or cyclist, then an effective test of  detection AND RESPONSE to ped/cycle road users will have been accomplished.  We could, even as laypeople, conclude that it works and it's adequate. System response cannot be correct if object detection is not correct.

As an alternative, a "naturalistic driving test" may be performed.  This method requires the vehicle to carry an independent system to observe all of the objects on the road, locate them, categorize them, and then try to verify which object the car was actually responding to.  More challenging, but possible.  In development manufacturers must do something like this.

While OEMs can indeed measure actual sensor performance such as target detection, ranging, tracking, categorization, and identification (what an "Eye Test" should show), they and their suppliers would regard those data as proprietary and not for public release.  At least this is my opinion based on nearly 25 years in the automotive industry and a total of 40 years in American industry.

I have deep background in automated automotive features for adaptive cruise control, forward vehicle collision warning systems, and forward vehicle collision mitigation systems among others.  My work on these convenience and safety systems spans from writing product requirements, testing and selecting actual sensors for development and for production, writing international standards (ISO standards) to set minimum limits for their performance, and supporting the developmental testing and refinement of a product line of such systems.  I was an appointed member of the team of US Experts selected to represent the United States in these proceedings, and for 8 years the lead author of the Forward Vehicle Collision Mitigation Systems standard, ISO 22839:2013.  These concepts for what can and cannot be tested publicly on production and production-like vehicles were developed and accepted by the International committee (ISO TC204, on which I served), as far back as 1998. 

I feel our request to the Senate would be much better received in the automotive community if we changed it to request a guarantee that appropriate performance requirements for the protection of pedestrians and pedalcyclists shall be written, implemented, tested, and verified for all production vehicles and all road-besed development vehicles.

I have not completed my review of S.1885, but my reading so far suggests that the needs I am trying to articulate are already provided for by the structure of the S.1885 draft you shared with me.  I hope to confirm my interpretation this week.

There should be a Docket open for this bill, in which the public could submit comments and see the response to them.  Are LMB or LAB using this avenue to provide responses to lawmakers?

Best regards,

Ken Freeman

Member of WBWC

Ann Arbor, MI