#21275
Postby jdlemon » January 8th, 2017, 5:40 pm
This is a really interesting report and deserves detailed reading, although I think the modelling is relatively limited. The headline negative travel time highlighted was a mere 1% in the Executive Summary, which I regard as negligible, so lets ignore that.
However Section 2.2.2 states:
"Behaviour at junctions and gap acceptance
Vehicles must identify a suitable gap in order to move between traffic streams, whether this is a simple lane- change, or a common conflicting movement such as a motorway merge or priority junction. The benefits of connected and autonomous vehicles may be to reduce this level of gap acceptance and better enable cooperative behaviour between vehicles on conflicting paths. Conversely, other work has suggested that10 technologies to assist in motorway or expressway driving (such as ACC) may impact capacity in merge or lane-drop situations, creating a bottleneck. This brings out a clear point; there is likely a trade-off where technologies designed to assist the driver and improve the driving experience in a given situation may negatively impact operations in another situation.
Research has indicated the potential for better provision of data – i.e. through connected vehicles – can reduce delays by encouraging early merging at junctions. Conversely, there is some evidence that automated vehicles behaviour, especially when pulling away at a signal junction, may reduce capacity. This is particularly the case if the behaviour of the vehicle, reflecting the preference of the user, is designed for comfort and safety, rather than traffic flow and road network capacity. Le Vine et al investigated the interaction between user experience and capacity at a signalised intersection. Assuming the level of comfort required to be the same as experience on high speed rail, reductions in capacity of between 21% and 54% were shown (at 25% fleet penetration). This work does not assume connectivity between vehicles, which may be of key importance; if a CAV has to assume a human driven vehicle may unexpectedly decelerate at its maximum rate, the requirement for large headways may naturally follow."
I think this identifies a huge potential hole in capacity terms at junctions until we have very high penetration levels.
For some time I've been mulling over traffic behaviour at a number of local junctions around Tunbridge Wells in slow speed or stationary traffic. In practice users on roads with clear priority from a legal sense have been allowing a considerable percentage of merging from a side road. I would anticipate that autonomous vehicles will have trouble replicating this behaviour and therefore rendering the junctions unnavigable from the side road for these vehicles and causing a change in traffic flows, presumably detrimental, although I haven't undertaken the modelling!
On the basis of this anecdotal evidence I think we've got a major problem with autonomous vehicles in the UK which has been generally unrecognised. This may not apply in the USA of course, where what we think of as good manners at these junctions may not apply and therefore not be included in appropriate test environments.