Section: New Results

Time-bounded message dissemination in strings

Participant : Gérard Le Lann.

In 2015, besides reviewing prominent open issues regarding safety in IVNs (see [42] ), we have investigated coordination problems that arise in string formations. Since the inception of the platoon concept (1977), a number of solutions have been proposed for achieving string control (platoons are a particular case of ad hoc/open string). String control must be exercised in order to avoid rear-end collisions, string instability, and for coping with emergency situations. The cyber components essential for string control have not been fully identified yet. For example, considering the cooperative adaptive cruise control paradigm, data collected in recent platooning experiments show that it is inappropriate to rely on V2V broadcast from a lead vehicle, thus the quest for other approaches. In strings, one can take advantage of short-range directional antennas which enable fast messaging among consecutive string neighbors, leading to the concept of neighbor-to-neighbor (N2N) communications and the cohort construct (a cohort is a string with a specification). String control problems translate into communication protocol issues and distributed algorithmic problems, notably:

• Time-bounded string-wide acknowledged message delivery and dissemination (TBMD),

• Bounded channel access delay (BCAD), a MAC-level problem,

• Time-bounded message acknowledgment (TBMA).

Acceptable solutions shall achieve small non-stochastic worst-case channel access time bounds (BCAD) and bounded delays for successful message delivery (TBMA and TBMD), under worst-case conditions regarding channel contention and message/acknowledgment losses. Non-stochastic worst-case bounds can only be established analytically (obviously, simulations cannot be considered). The importance of the TBMD problem can be exposed simply as follows: would TBMD be solved, then the string instability problem vanishes. Rather than resting solely on stepwise detection-and-reaction strategies based on radars/lasers, every string member adjusts its acceleration/deceleration rate according to observed motions of its predecessor, TBMD delivers a N2N message carrying the newly string-wide targeted velocity, in less than 100 milliseconds in strings comprising in the order of 20 members, in the presence of message/acknowledgment losses. The TBMD problem has been solved (see [34] ). The solution rests on assuming that TBMA and TBMD have solutions. Both problems have been solved (solutions are under review). Contrary to strings, groups are ad hoc/open multilane formations. It turns out that solutions aimed at the 3 problems referenced above are instrumental in solving problems arising with multilane SC scenarios. For example, the 3-way handshakes at the core of safe lane changes published previously now achieve significantly better performance figures. Work in progress also includes:

- conflicting concurrent lane changes at high velocities,

- fully automated zipper merging at high velocities, in non-line-of-sight conditions (radio communications), in line-of-sight conditions (optical communications).