The Nuclear Bunny Blog

I have two different late model (a 2006 and a 2007) GM vehicles with satellite navigation systems. Both are manufactured for GM by Denso, although they clearly have different implementations and features. The one thing they both share, however, is blatant stupidity when calculating routes.

I travel quite a bit between Houston and Austin and there are really only three reasonable ways to go: US-290; I-10 & TX-71; or I-10 & TX-183. Over the years my preference has moved from US-290 to I-10 & TX-71 as traffic along US-290 has increased on the eastern fridges of the Austin area, but the distance and time between the two is very similar and really it depends upon where in Austin you might be headed.

The GPS systems in my cars have three route options when you pick a destination: fastest, shortest and other. In both cars, the route displayed is none of the above three, but instead a long, out-of-the-way route of I-10 to Seguin before heading north on TX-123 to San Marcos, and then heading into Austin on I-35. This option is a whopping 222 miles with an estimated travel time of 3 hours and 39 minutes.

Now, the fastest route is usually not the shortest. Route calculations are supposed to take into account the actual speeds of the road segments involved, and since interstate highways are almost always faster than anything else, there is a natural preference to these roads. In this case, however, the calculation is way off. Taking I-10 to US-71 (via US-59 & TX-8 inside of Houston for those familiar with the area) is only 166 miles and estimated to take 3 hours 29 minutes according to Google. Likewise, taking US-290 is only 168 miles and estimated to take 3 hours 22 minutes.

Of course these time estimates are usually worst case scenario. Driving a few miles per hour over the speed limit I often make my trip in 2 hours and 30 minutes; worst case has been 3 hours and 10 minutes when driving 55 as an experiment (a painful one at that).

What about the shortest route option in the GPS system? That does actually route you onto I-10 & US-71 in this example, but the shortest route calculations are always incredibly stupid, too. They will often take you onto side streets for half a mile or less just to maintain the shortest possible distance. Once, when driving through Memphis, TN, the GM GPS took us on and off the interstate highway three times when driving through town, often for just a couple of blocks, just to maintain the shortest distance. Granted this is a clear example of a GPS system not being there to replace your own intelligence, but if you don’t know the town…

In the early days of consumer-level GPS systems, one of the common problems was having accurate speed data for different road segments. In western states especially, secondary highways often have the same speed limits as interstate highways. In Texas, the past 20 years has seen many of these highways improved so they route around the smaller towns, rather than going through them and stopping at every light. Consequently it is often better to take these roads than the interstates whenever possible. If the GM GPS systems do not have accurate speed data then it can account for some of this behavior, but not all.

On my normal route, once you finally hit TX-71 and the GPS decides you mean it, suddenly the estimated arrival time drops dramatically along with the distance. Is it computing this arrival time based upon your actual speed so far, or based upon the speed data of the road?

Over the years I’ve gone from absolutely wanting a satnav system in the car to wishing mine didn’t. I get much better results using a smart phone with Google Maps to get a general idea of the route to take and then consulting with it now and then if things get tricky. But for the most part, they seem poorly done enough that they just get in the way rather than help. Will the newer generations out now that provide real-time traffic data and have hard-disk based data improve their logic and performance enough that these problems are solved?

One final note for embedded systems developers: you don’t get to break the interactive response time rule of 250ms just because you are writing code for an embedded system.

I recently purchased a Garmin Forerunner 310XT training device for use while running, cycling, and hopefully swimming. The Forerunner 310XT is a new device from Garmin, and their first multi-sport device that is waterproof and can be used for swimming, and thus triathlons.

For the past several years I have been using a Polar S725X multi-sport training device for both running and cycling. The rest of this review will compare and contrast to the Polar device since that is what I am most familiar. I have not used other GPS-based training devices before, so comparisons with those are left as an exercise for the readers.

The 310XT uses GPS to calculate your location and speed. You can then upload your data via your computer to Garmin’s Connect website, Garmin’s Training Center software, or other third-party options. After uploading you can view a map of your route, categorize and describe your activity. With the Connect website, you can easily share your activity via variety of methods, and even export the route into Google Earth.

Like most of Garmin’s training devices, you can pair the 310XT with a variety of other devices that communicate using the ANT+ Sport protocol. The 310XT will receive and record data from heart-rate monitors, bicycle speed & cadence sensors, foot-pods, and even power meters such as the Saris PowerTap. The collected data is then joined with your route data and transmitted along with it.

The bicycle sensor option from Garmin is a combination speed & cadence sensor that mounts on the rear chainstay. This design allows one sensor unit to have two separate magnets for both the crank and the rear wheel. Compared to the Polar solution, this is much cleaner and easier to mount than two separate sensor units. The Polar cadence sensor, especially, is difficult to mount on some of the modern carbon downtubes, so the Garmin solution is a welcome change.

The speed sensor augments the GPS data so that accurate speed is recorded even when GPS signal is not available or accurate.  Garmin also sells an optional foot pod sensor that accomplishes the same when running without GPS signal, such as inside on a treadmill. I haven’t used this sensor yet as where I have run has had great GPS reception thus far.

I used a foot pod sensor with my Polar to keep track of my speed and distance while running. The downside of this type of solution is the relative inaccuracy of the data. I’d often see 10 to 20% margin of error, even after calibrating the foot pod.

My concern with a GPS-based training device would be the accuracy of the signal, especially when running under under a lot of foliage. So far, the 310XT has proven itself more than capable in this regard, and much better than other GPS devices I have used in the past. Viewing the recorded route data I’ve found very little error in position so far, and only a little bit with elevation, even when running underneath trees and bridges.

Like all GPS devices, the 310XT must determine its position when it is powered on. If you start up the 310XT inside, or where there is poor GPS signal, it can take a very long time to determine initial position, if it all. This can be an annoyance when you’re ready to start your training and you don’t have a solid 3-dimensional fix yet. The 310XT lets you begin your timer and start your activity before it has GPS signal, but it can sometimes take surprisingly long to get that initial fix. Another disadvantage of not having a fix is you won’t have accurate local time until then, either.

The Garmin 310XT has a rechargeable battery, but battery life is only approximately 20 hours of usage. My Polar S725X, by comparison, has a watch-style battery that has lasted over 5 years. In practice, a rechargeable battery solution works great, but may present a problem on multi-day activities where there is no opportunity to recharge.

Transmitting the data from the training device to your computer is done via a USB dongle that uses the ANT protocol. Garmin provides drivers and software for both PCs running Windows and Macs running OS X. The Polar solution uses an infrared based receiver that is much more difficult to work with as it requires line of sight between the training device and the infrared receiver, and Polar does not provide a software option for Macs.

For most users, Garmin’s software solutions are stellar compared to Polar’s. The big win here is Garmin’s Connect website which allows you to easily view your data online, share it with friends, and upload your data from a variety of different computers. Sharing your data is trivial, and allows your friends to view your routes, as well as being able to search for routes from others near you. The Connect website represents a modern solution for the social Internet, a great solution for most users.

Where the Polar software shines is for advanced heart-rate based training, especially with coaching assistance. Polar’s software is designed to easily share your data with a coach and receive training programs with them. Additionally, both the software and the Polar device have much more advanced features for heart-rate based training. For example, with the Polar I can run several tests to determine if I am over-training, or even what an estimate of my current VO2max. The Garmin solution only has different heart rate zones, more than adequate for most users, but you do get the impression the Polar has more science behind it.

The 310XT supports multiple bike settings, each with its own odometer. I find this feature in particular compelling as I like to keep track of my mileage for each bike separately, and as far as I could tell the Polar S725X always combined bike mileage for all bikes it had configured, rather than a separate odometer for each one. With the GPS based tracking, you don’t even necessarily need a speed/cadence sensor and you can still keep track of your bike’s mileage. This is especially compelling for mountain biking, where the rough trails often render a traditional bike computer useless.

So far, the Garmin Forerunner 310XT has been a fantastic training tool and I haven’t missed using the Polar S725X yet.

I’ve been a GPS user since the system was first available and one thing I have always wanted was the ability for navigation software to take additional factors into account when planning a route. Most systems just offer you the ability to build routes based upon road speeds, road types (highways, toll-roads, off-road, etc.), and a choice of speed or distance goals. That’s OK, but any seasoned travel knows there’s more to it than that.

What about avoiding known traffic bottlenecks? Some navigation systems started adding that a few years ago. But one that I’ve been waiting for forever is the ability for a navigation system to route you around bad sections of town. If you are from out of town, you likely have no idea what sections of town to avoid. All the data a navigation system needs for that are already out there – crime rates by ZIP code, for example.

Honda now has a navigation system that does exactly this. Unfortunately, it is available only in Japan at the moment.

I’m sure in the overly PC society that the Western world has become it will never find its way here, which is a damn shame.