 |
| 2nd Gen. Tesla Motors "Autopilot" Perception IO |
Falling Short of Glory
Sleepy, running late, confused, on drugs, angry, emotionally disturbed, distracted, people can be great at a driving but often fall short of their potential because of the aforementioned incomplete lists of reasons. Our human senses also fall short of that synergy of Radar, Lidar, Machine Vision & Artificial Intelligences that autopilots vehicle will make use of with consistent, superlative safety control far exceeding that of even the best human driver by an entire order of magnitude or better!
Human Computer Racing
While humans have been racing cars, recently computers have been doing it with even greater precision. I wonder about the purpose of racing cars in reality if the people have been removed from the driving equation. Training to become a professional racer take a lifetime dedication starting in shifter car racing to get to the Formula 1 or F1 level. Taking people out of the driver seat might seam a little anti-human until you start to think about safety and the fact that it takes a lot of smart people working as teams to make self driving cars race each other. Learnings from these kinds of races can be ported to consumer self driving car autopilot control systems for "track day" fun as a passenger in your Tesla Model S in "race driver ride along demo mode"--- a pipe dream only eclipsed by a world of Hyper-loop enhanced Pop.Up drone train car hybrid networks of vehicle sharing that Airbus just started promoting.
Learning as a Team!
When a person makes a mistake while driving, all they can do is commiserate their mistake, perhaps at most sharing with others in a post online or in a conversation. Sadly the learning is not widely adopted by other people. When the machine controlled vehicle makes a mistake, the mistake is shared between all the machines on the network, thus all current and future autopilot vehicle learn as a team, as one big connected data exchange system of shared machine learning. V2V or vehicle to vehicle data exchange will allow connected vehicles to talk to each other. V2G or vehicle to grid will allow connected vehicle to talk to infrastructure, like intersections control systems.
Ehang 184 & Pop.Up. Airbus
Flying cars like the Airbus Pop.Up. drone lift + wheeled skateboard pod + train pod Hyper-loop may be grabbing the headlines lately, but a few years ago Ehang was the stud of the show with its 184 short hop double safe autopilot drone vehicle. Electric flying machines that hop over traffic in megacities above the tail pipe pollution & congested road traffic, these are going to be the future of luxury personal freedom transportation solutions.
Fully Electric Tesla
As we continue moving into the future with electric cars, more information technology will be integrated by automakers for the enhancement of safety, efficiency, system analytics, emissions controls, driver assists, and other extended functionality granted by innovative technology integration. Today the Tesla Model S leads the way!
Hybrid Synergy Drive
Back in 1997 executives at GM criticized Toyota for developing the "hopelessly" complex Prius family of vehicles. At the time it was believed within the upper ranks of GM that building a car with that many actively computer controlled systems could never be done in a fiscally efficient or reliable way, especially not on a commercial scale. Proving GM's leadership wrong, Toyota has churned out more than 10 million hybrid vehicles. Toyota demonstrates beyond a shadow of a doubt that highly complex vehicles can be produced in large volume without compromising legendary Japanese reliable endurance properties emblematic of the Toyota Prius.
Skip OEM Sat-Nav
OEM satellite navigation systems are always slow buggy half baked examples of what you can find in a recently released smartphone, stand alone Garmin GPS unit or alternative aftermarket navigation solution. To be fair, the OEM Sat-Nav systems are expected to hold up in the car for many years longer than the average life of a smartphone.
Mental Navigation
The temporal lobe of your brain can be used for navigation too, once you are well familiar with a route or area! YMMV depending on neurological functioning level, general IQ, sensory performance of your vision and hearing as well as motor control of your limbs.
Increasing Complexity
The complexity of automobiles will continue increasing as more complex digital hardware is integrated for feature enhancements, safety improvements, efficiency improvements, information connectivity, on-board diagnostics, and other IT related system enhancements.
Self Driving Future
Autopilot the pinnacle of information technology integration into vehicle designs. A mainstay in aerospace for decades, autopilot was easy to integrate into capital intensive vehicles like commercial aircraft, where the cost structure of the autopilot represents only a small percentage of total vehicle costs. Once in flight, the hazards to aircraft are less complex than those encounter on roadways by passenger cars, trucks and other wheel road vehicle transportation systems. Load up youtube with some dash-cam fail videos to see examples of how even self aware people with all our intellectual gifts are able to turn a driving situation fubar in fractions of a second from one bad decision.
Seeing in More Ways
Computers that drive cars will have a sensory advantage over human eyes & ears, so much so that the self driving car can see in 360 deg with hundreds of frames per second, using a combination of lidar, sonar, radar, machine vision cameras, kinetic analysis, environmental monitors, and other types of emergent sensor input technologies like GAMA & FLIR cameras.
General Intelligence
Humans have one huge advantage over autopilot system today. Within the 700 TB of memory computer comprising the adult humans general intelligence, humans are able to quickly recognize a bewilderingly complex array of different things using neurological information processing techniques that are blisteringly fast by comparison to Van Neumann digital information processing architectures. The only way computer controlled cars will be able to keep up is through network connected system learning, software updates, and ongoing optimization. Since the cars will be able to talk together as a team faster than people can communicate with words, the networks of autopilot vehicles will quickly eclipse the cognitive performance of an individual human driver.
Recursive Optimization
Future generations of autopilot vehicles will build on the advances made in previous generations, eventually greatly surpassing anything that a person is able to do as a driver in terms of vehicle operation efficiency or safety. In fuel powered, hybrid, and electric vehicle, autopilots will always achieve greater range with less energy per mile, lower emissions & continually improving safety! Once singularity is achieve the autopilot will also become more intelligent than people.
GPU to Drive You
In one generation of GPU design architecture, Nvidia was able to use AI to radically improve performance while also reducing power consumption substantially. Nvidia is partnering to bring some of their GPU tech & A.I. technology to automakers. We know that the news of these partnerships today means than models released in 2021 will make use of Nvidia technology.
Interdisciplinary
Drug discovery with Quantium Computers might sound unrelated to self driving cars until you consider the big data similarities between custom therapeutics, genomics, informatics, artificial intelligence, machine learning, biochemistry, physics & astronomy. Even more promising is the Neurological-Synaptic "True North" processors under development at IBM for the miniaturization of Watson into a 60w pizza box sized machine. These same "neuron" style integrated circuits are able to do heavy data analysis in the Teraflop range using tiny amounts of power. A future 10w peak cell phone making use of a synaptic chip set will have the computational performance peak of a small server farm of van-neuomann cloud of racked blades with high bandwidth optical bus connections.
Laser Vision
BMW may have the coolest laser headlights, but since 2003, their vehicles have been plagued with long term reliability problems rooted in electrical glitches, shorts, sensor failures, ghosts in the machines and other hard to solve electrical problems that can be mostly blamed on BOSCH! I have been rooting for BMW ever since learning about the Carbon Fiber Electric I3 & radically awesome I8! Woot woot to efficient dynamics ^^
No comments:
Post a Comment