Inside GNSS Media & Research

SEP-OCT 2018

Issue link:

Contents of this Issue


Page 24 of 67

S E P T E M B E R / O C T O B E R 2 0 1 8 InsideGNSS 25 with combining multiples sensors and systems in ground vehicle applications, overall system connec- tivity, simulation and road-testing implementation challenges as well as other navigational sensors that could augment ground vehicle applications. RANGE ASSOCIATION Spenko opened the autonomous vehicle navigation discussion with his "pieces of the safety pie", which include: path planning, mechanical/electrical, con- trol and navigation. For the purposes of this webinar, Spenko focused on the navigational component. Fundamental to the autonomous process is to de- termine a robot's location in space based on all sensor input while accounting for sensor noise and undetect- ed faults that may exist in those sensors. Navigation faults come from raw sensors (e.g., GPS, IMU) or range-based sensors (e.g., LiDAR, radar, cameras). Sensor noise is a well-established field of study. Spenko explained, "What we don't know how to do well is account for undetected errors and faults— those faults that may occur in our [range-based] sen- sors that we haven't been able to detect and remove." Range-based sensors don't provide an actual position, but there's still a need to extract features from that data and associate that data with a land- mark. Spenko added, "We spend considerable time trying to analyze how these algorithms introduce faults into our navigation safety." While other research groups have used experimen- tal or simulation-based studies to evaluate algorithms, Spenko and his team have shifted direction, explain- ing, "The problem [with experimental methods] is that anytime you want to make change to algorithms or update sensors, you have to restart [the test] clock." Instead, the IIT group is taking some inspiration from the aviation industry to gain a measure of trust in sensor information. Of course, there are signifi- cant differences to aviation and terrestrial robotics: 1) GNSS alone is insufficient. Terrestrial robots will en- counter a number of GNSS-denied areas; 2) Aviation integrity risk peaks at landing, terrestrial robots need continuous risk monitoring; and 3) Aircraft have con- trolled runway environment; mobile robots live in a dynamic environment that changes with people, cars, cyclists, construction, etc. AT ATTENDEES ANTED TO KNO Participants had the chance to ask questions during the webinar. Here are a few of those questions: •   Now that we are seeing all these different sensors that are  going to be integrated into an automotive system, do you  think this is going to change the relationship between OEM  sensor suppliers? Do you think it's going to lead to more  vertical integration for example like we see in the cell phone  market as opposed to horizontal integration? •   Tell us more about absolute sensors pushing as a solution. •   In addition to the three key differences noted between  aviation applications and ground vehicle applications, can you  comment on more differences? •   What kind of testing facilities or kind of features in testing  equipment do you require for testing out these autonomous  navigation systems? •   How do you scale the Locata system in a cost effective way?  Or how does it scale up? •   How should you characterize errors in the integrity  analysis that you are doing for these autonomous vehicle  applications? That is, in regard to artificial intelligence, how  do you model errors? •   Is it feasible to extend and create these HD maps for urban  areas where RTK availability may be intermittent? If not are  there alternatives to 10 centimeter and 95 percent confidence  in urban areas? •   With the increase in GNSS constellations do you ever see a  case where it's going to be possible to do these GNSS-only  navigation solutions and use other sensors just for integrity  monitoring? •   Are there established performance standards, metrics that  are considered by the automotive companies while testing  their systems? •   Can you say something about other RF based positioning like  cell phones, etc.? How does this compare and/or compliment  Locata solutions? •   In terms of the poll question dealing with experimentally  validating the performance, how much data collection is  required to validate one of these systems? ADDITIONAL SUPPORT •   Watch a CAV Testing—Tools Integration video: https:// •   Connect with Spirent via linkedin: showcase/spirent-positioning/ •   Visit Spirent online at: TO WATCH THE FULL WEBINAR, GO TO: https://register.

Articles in this issue

Links on this page

view archives of Inside GNSS Media & Research - SEP-OCT 2018