To GPS or not to GPS?

To GPS or not to GPS? The self-driving car question

Most self-driving cars are expected to get by with a combination of radar, lidar, and cameras to scan the world around them and move through it safely. This process shows considerable potential for moment-to-moment driving, but it runs into limitations when it comes to navigation, understanding where to drive on unmarked roads and ensuring autonomous vehicles know exactly where they are at any given moment. This is where global positioning systems come into play, but considerable innovation is needed to get GPS solutions up to the task of supporting self-driving cars. As GPS technologies evolve, the industry is left wondering if they should pay heavily for the kinds of systems necessary or hold out for alternative options.

Using GPS for self-driving cars

In a typical navigation environment, a GPS need only be accurate within a range of a few meters. In a self-driving car, that type of precision can lead to cars making an incorrect evaluation of where they are on the road relative to the navigation systems present and make a mistaken move. GPS solutions for self-driving cars must be accurate to within a range of centimeters, and they must provide that accuracy at all times, without the disruption caused by lack of visibility within satellite or base tower networks. Bringing together its own analysis with information from ABI Research, an SAE International report explained that next-generation GPS technology is beginning to emerge and will likely become fairly mainstream by 2021.

The most accurate GPS systems currently on the market will use a combination of a base station and a rover to perform surveys using dual-frequency technologies and being accurate to within 1 cm horizontally and 2 cm vertically, Todd Humphreys, associate professor at the University of Texas at Austin, told the news source. As of now, those systems typically cost anywhere from $1,000 to more than $2,000.

Establishing a network of base stations, rovers and specialized GPS setups in self-driving cars could prove incredibly costly, and precise point positioning – an alternative option that works if there isn’t a base station nearby – can require minutes to get an accurate reading. Humphreys explained that achieving the goal of next-generation GPS requires improvements in cost efficiency and a significant reduction in the amount of time it takes for the platform to calculate and gather location data from raw readings.

While these conclusions coming out of the survey-focused GPS industry are serving as a guiding point for innovation in self-driving cars, a few brands are leading the way on GPS architectures that are more viable in autonomous vehicles and accurate within a 5 cm range. This charge is being led by EXO Technologies.

How EXO Technologies is changing the GPS game

Predictive technologies are at the center of the EXO Technologies model. Advanced satellite ecosystems are used to gather and create GPS data, while predictive technologies identify any possibilities of error within the calculation. This can include precision problems caused by atmospheric interference, clock issues or ephemeris. Predictive analytics pin down the possibility for error to provide a first fix time of under 1 second, a dramatic improvement over what has typically taken 20 minutes or more. With accuracy down to approximately 5 cm and no dependency on base stations, EXO Technologies is laying the groundwork for a simpler infrastructure model surrounding the GPS, allowing for lower total costs.
Details on the precise pricing and technical makeup of the EXO Technologies PICO GPS ecosystem are unclear as the company is still in an early startup phase, but it is clear that the solution is directly targeting the overarching costs of maintaining a GPS system by eliminating the dependence on base stations and rovers without making a major sacrifice to accuracy. This strategy also falls in line with the overarching tech industry trend of moving solutions over to software – in this case using predictive analysis to boost accuracy – allowing for easier scaling of systems instead of depending on dedicated hardware to resolve technical problems.

Next-generation GPS offers a powerful option for self-driving cars, but it may not be the only opportunity out there. There is a growing movement to improve geospatial mapping in autonomous vehicles to reduce the burden placed on GPS solutions.

The potential of mapping and leading players

In theory, a vehicle will be less dependent on a GPS to tell it where it is if it can be more aware of its surroundings and compare those readings with accurate maps of the area where it is driving. This type of mapping represents the fusion of machine learning and crowdsourcing, as vehicles can share data about what they see to create detailed digital maps that vehicles can use to route an area and learn the best way to move safely to a destination. Of course, this represents a huge virtual learning curve, as detailed driving data is required to complete these types of maps. At this stage of the self-driving car sector’s evolution, leading players in the segment include:

#Civil Maps

Civil Maps aims to focus on self-driving vehicle cognition to give cars a better idea of where they are and what is going on around them. This includes using an augmented reality map for navigation purposes and crowdsourced reference data. The solution currently focuses on using lidar to perform environmental analysis.

#Here

With a focus on moving beyond robotics, Here is working to use dynamic maps to help autonomous vehicles proactively analyze the environment around them and make safe driving choices that are human-like using high-definition virtual maps that allow vehicles to interpret the area around them with centimeter-level precision.

#Oxbotica

This UK-based startup is taking the idea of real-time digital mapping to another level by using two stereo cameras alongside laser scanners to create a digital three-dimensional map of the area around a vehicle. With this map in place, the vehicle can move around and see the environment change, allowing it to use a 3D point cloud ecosystem to drive safely, Popular Science reported.

Moving ahead

Choosing between a next-generation GPS solution and a digital mapping setup is, most likely, a project-specific decision. Mapping requires a great deal of data gathering, something that will be much easier when there are more self-driving cars on the road to do the crowdsourcing work the industry needs. Solutions like the Oxbotica model could change that, but GPS is already showing a degree of readiness that may make it an attractive option for those that want to get vehicles on the road soon. Ultimately, however, the decision may end up being much more like typical technology choices and not become an either-or-decision, but one where both technologies end up working together.