The buzz around driver-less cars is hotting up. Suddenly Google are being made to seem slow-to-market by Tesla and Uber. It seems the machine-driven car is here already and aside from a few years of human-aided operation, 'the machines' will have a big impact on our modes of transport.
I never understood why trains needed drivers in the first place. They run on tracks, the junctions are controlled by points operated by machines, and there is minimal traffic to worry about. I suspect it may be customer fear and major reluctance from laggard trade unions that have restricted the automation of what seems to be a much simpler transport problem. At least we are seeing a gradual introduction on newer lines (Docklands Light Railway) and some older lines (Victoria line), albeit with human 'train captains' still on board to handle emergencies.
The easier train lines to automate are new ones, which don't suffer from the same key problem as driver-less cars do. The problem is humans. Wherever automation and humans need to coexist there is more complexity and a much higher degree of risk.
This is the key problem when putting driverless cars on the roads. It's not the computers that are the problem but the illogical, unpredictable, and frankly hazardous humans.
On 30th June 2016 Tesla reported that it had made the NHTSA (National Highway & Traffic Safety Association) aware of a fatal accident. A Tesla Model-S collided with a tractor trailer on a divided highway as the tractor was crossing perpendicular to the road. The Tesla autopilot failed to identify the trailers white side panels against the sky and did not detect the chassis of the trailer. This tragic event left a family without a father and served as a stark warning to the driver-less car industry.
Even in countries with established and, by-and-large, good adherence to road safety and conduct rules, there are between 5 (UK) and 12 (USA) deaths per 100,000 vehicles. that works out to 3 (UK) and 7 (USA) fatalities per billion km's driven. The Tesla fatality (US) was the first in over 200 million km's, indicating a better safety level than that of human-driven cars.
In India 130 people die per 100,000 vehicles. In Bangalore, where I find myself this week, 2 people die on the roads every day. I wish I'd have known this figure before I rented a motorcycle for the day. I might not have been so keen.
The reasons for this alarmingly high death toll are numerous. There is appalling traffic congestion, there are overcrowded and poorly maintained pavements, there are no traffic lanes, huge numbers of scooters and small motorcycles. The clearance between vehicles is often inches rather than feet. Nervous drivers go nowhere, and traffic signals are like the pirates' code - more guidelines than rules. Add into the mix a few thousand tuk-tuk taxis, poorly maintained vehicles and roads, animals running loose in the road, and you need a vastly more complicated compute model for driver-less vehicles.
If the Google/Uber/Tesla software wants a real test, they should be (and probably are) testing here in India. If they can drive us safely in Bangalore, Mumbai and Delhi, then Los Angeles and London will be child's play for 'the machines' to navigate safely.
I never understood why trains needed drivers in the first place. They run on tracks, the junctions are controlled by points operated by machines, and there is minimal traffic to worry about. I suspect it may be customer fear and major reluctance from laggard trade unions that have restricted the automation of what seems to be a much simpler transport problem. At least we are seeing a gradual introduction on newer lines (Docklands Light Railway) and some older lines (Victoria line), albeit with human 'train captains' still on board to handle emergencies.
The easier train lines to automate are new ones, which don't suffer from the same key problem as driver-less cars do. The problem is humans. Wherever automation and humans need to coexist there is more complexity and a much higher degree of risk.
This is the key problem when putting driverless cars on the roads. It's not the computers that are the problem but the illogical, unpredictable, and frankly hazardous humans.
On 30th June 2016 Tesla reported that it had made the NHTSA (National Highway & Traffic Safety Association) aware of a fatal accident. A Tesla Model-S collided with a tractor trailer on a divided highway as the tractor was crossing perpendicular to the road. The Tesla autopilot failed to identify the trailers white side panels against the sky and did not detect the chassis of the trailer. This tragic event left a family without a father and served as a stark warning to the driver-less car industry.
Even in countries with established and, by-and-large, good adherence to road safety and conduct rules, there are between 5 (UK) and 12 (USA) deaths per 100,000 vehicles. that works out to 3 (UK) and 7 (USA) fatalities per billion km's driven. The Tesla fatality (US) was the first in over 200 million km's, indicating a better safety level than that of human-driven cars.
In India 130 people die per 100,000 vehicles. In Bangalore, where I find myself this week, 2 people die on the roads every day. I wish I'd have known this figure before I rented a motorcycle for the day. I might not have been so keen.
If the Google/Uber/Tesla software wants a real test, they should be (and probably are) testing here in India. If they can drive us safely in Bangalore, Mumbai and Delhi, then Los Angeles and London will be child's play for 'the machines' to navigate safely.
No comments:
Post a Comment