15 replies to this topic

[maxay1]

My sincere apologies to Dave Williams: I think I have abused your kindness and your generous nature with my innumerable questions and queries....but my interest in active suspension and the promise it holds has gotten the better of me. So....yes, I have started this thread with the full expectation that the genius behind active suspension might be willing to contribute to it ("it" being the thread).

The implications it holds for suspension design and the resultant kinematics are significant; camber gain and toe-steer in roll and pitch would be greatly reduced, given the ability to keep the platform 'flat'.

I thought it might be interesting to discuss the technical development of active suspension, how it addresses the 'ride/handling compromise', stability and control, and what changes might be made to a modern-day implementation of active suspension with the benefit of experience.

Thanks in advance to all who contribute (constructively)....

Wil

[DaveW]

To my knowledge, two types of active suspension were developed for use in F1. The Lotus system was irreversible, in the sense that the "corner" springs & dampers were replaced by four servo-controlled hydraulic actuators. The "suspension" actuators moved only when commanded to do so. The other, a hydro-pneumatic system loosely similar in principle to the Citroen system, first appeared on the Williams vehicle (I think), developed, or at least mentored, (again I think) by Frank Durney. A hydro-pneumatic system will respond naturally to road, aero and inertial inputs, & the system must work hard to minimize undesired responses.

I can talk about the first type, having been involved closely from the initial brain storming throw away, to the first attempt in 1983, and through to the point where the 99T became the first actively suspended vehicle to win a Grand Prix in 1987. The episode was, for me as an aeronautical engineer, a considerable learning experience at several levels. It was also a huge privilege to work closely with the very best technical people (including automotive, hydraulic and instrumentation engineers) and drivers along the way.

Our objective was to design a control system that responded only to road inputs, and to develop the concept on an Esprit "buck" that generated no down force (to speak of). It seemed at the time to be unambitious to emulate a conventional suspension, so we implemented a "modal" suspension, conceptually comprising heave, pitch , roll & warp springs & dampers. I have to admit that I omitted the warp compliance initially, such was my lack of knowledge. Bruce Maclaurin of RARDE soon pointed out that it might work better if it was included, & he was right. The concept proved to be a good one - at least it remained unchanged for as long as I was involved - but it did raise a whole series of questions that could be resolved best by thinking from first principles. Some examples are:

- Is it necessary to allow some roll with lateral acceleration?
- How is the lateral mechanical balance to be adjusted when the vehicle doesn't roll? Answering that lead naturally to the idea of an algorithm to maintain a neutral balance under most conditions (even mixed snow & dry tarmac).
- Is it necessary for the vehicle to squat under acceleration?
- Is it necessary to change the mechanical balance under acceleration?
- Is it necessary for the vehicle to pitch under braking?
- Is driver "feel" important.

Edited by DaveW, 23 August 2010 - 12:14.

[maxay1]

Dave,

Thanks kindly for the reply, and your response raises some interesting issues regarding the implementation of active suspension. Particularly driver interpretation of vehicle response to control inputs. Chassis roll (velocity, not angle?) may be an important cue for the driver at larger lateral accelerations, but in the absence of it, the driver is still subject to inertial reaction, I suppose. This should give some clue to the driver, as well as any o/s or u/s characteristic built into the control algorithm.

It would be interesting to know how one would separate platform from unsprung mass and make changes to the lateral balance of the vehicle, especially taking into account surface irregularities, and CPL variation.

Thanks again for contributing.

Regards,

Wil

[StressedDave]

Our objective was to design a control system that responded only to road inputs, and to develop the concept on an Esprit "buck" that generated no down force (to speak of). It seemed at the time to be unambitious to emulate a conventional suspension, so we implemented a "modal" suspension, conceptually comprising heave, pitch , roll & warp springs & dampers. I have to admit that I omitted the warp compliance initially, such was my lack of knowledge. Bruce Maclaurin of RARDE soon pointed out that it might work better if it was included, & he was right. The concept proved to be a good one - at least it remained unchanged for as long as I was involved - but it did raise a whole series of questions that could be resolved best by thinking from first principles. Some examples are:

Now I'm getting all emotional... my first job out of university was as Bruce's underling running the active tank.

I'm fairly sure the Williams active system was born out of one of Bruce's other ideas for active suspension that he'd built in conjunction with AP. The prototype (another tracked vehicle) was sitting in a shed at RARDE when I was there. IIRC that one used mechanical accelerometers to control the various valves as when it was first built TI DSP chips hadn't been thought of.

Edited by StressedDave, 24 August 2010 - 07:05.

[bombastic]

I know of one "Lotus Senator", an opel saloon car with a fully active Lotus suspension being used at the technical university in Aachen. It was a hell of an efford to get the system working again, after being shut down for several years.

The last I know is that the car was used in a project about driver reactions and feelings to a chassis' characteristics as it is possible to change the behaviour from under- to oversteer or vice-versa in a single corner.

Or to let the car drive throu a pothole that doesn't exist. A really interesting project. But the suspension also uses a lot of power.

[StressedDave]

Early Lotus systems would do that as you had to drive the ram in both directions. Letting the vertical forces do that work made for large energy savings. The downside was the need for a pair of Moog valves per corner.

[Tony Matthews]

I'm fairly sure the Williams active system was born out of one of Bruce's other ideas for active suspension that he'd built in conjunction with AP.

The Williams system was a modified AP system.

[mariner]

Sadly it would not be allowed in F1 but the development of radar for cars makes one wonder if a combination of specialised forward radar to map the terrain and active would be interesting.

Clearly the radar could only look a limited amount ahead but if it was mounted on the roof (and maybe could sweep side to side from steering inputs) it could provide an external reference point for the active system to refer to when ordering ram movements. I suppose that is a bit analagous to the terrain hugging systems on combat aircraft. I read somewhere that the pilot can select the degree of terrain following to smooth the "ride" at the expense of sometimes flying higher!

One question about the Lotus F1 active system, IIRC it was later fitted with springs to assist in the pits and "wheeling around". I wonder if these provided much support at racing speeds.

Also if the Lotus 88 was illegal did anybody ever challenge active as I assume it could almost anything with ride height if desired.

[DaveW]

I know of one "Lotus Senator", an opel saloon car with a fully active Lotus suspension being used at the technical university in Aachen. It was a hell of an efford to get the system working again, after being shut down for several years.

Actually, I recall we built around 50 systems for various applications. Two of our F1 systems were, I believe, used to control the suspension of the Thrust SST.

I'm sure it would be a (commendable) effort to resurrect an unknown system. I'm impressed.

A really interesting project. But the suspension also uses a lot of power.

I'm pleased you found it an interesting project. I recall that a road vehicle system consumed, on average, about the same power as an Air con. unit. It is true that power consumption could have been improved, but our task (I like to think) was to demonstrate the possibilities of active suspension; the task of turning it into a "production" system it was the role of a small group set up by Bill Moog in Stuart, Florida. Incidentally, a passive suspension system also consumes power (dampers heat up when they are worked, & there is only one source for that heat).

Loads, number of actuators, & power/weight were quite different for the Scorpion, of course.

Many of our OEM customer engineers were seeking a "magic carpet" ride, and "preview" (forward looking radar) was one of the ideas suggested to achieve that goal. You might like to consider the following:

Statement: the velocity of an hydraulic actuator is proportional to the signal used to drive the EHSV, provided sufficient authority exists.

Now here is a thought experiment.

Measure the vertical acceleration of the wheel hubs. Integrate in real time to obtain vertical inertial velocity. Now add that signal to the EHSV drive signal. Each actuator will have an added velocity that can be exactly equal and opposite to that of the corresponding hub velocity, provided that the signal is suitably scaled & "signed". This idea can be used to "fix" the position of the sprung mass inertially in the presence of road inputs (i.e. no load will be transmitted by the suspension to the sprung mass) regardless of the hub input - i.e. perfect ride, at least until the actuators run out of stroke.

The thought is useful because it illustrates a concept. It can even be implemented, with a few additions, for some applications (e.g. an active truck cab suspension). However, hub motion will not be damped by such a control system (no load applied to the sprung mass = no load applied to the unsprung masses) so the concept, as stated, is flawed in the vehicle suspension application, although it can be (& was) used in moderation to manipulate the ride/handling compromise.

The experiment is, I think, a simple way of demonstrating that the holy grail of road car designers, a "magic carpet" ride, is obtainable only if vehicle control is abandoned. At a slightly deeper level, it also suggests that "preview" or "look ahead" is unnecessary except, perhaps, to optimize available stroke before encountering an extreme input. Distinguishing between bricks & paper bags is also something of a problem.

p.s. (to Mariner) Springs were fitted in parallel with the actuators on all the Lotus active vehicles (apart from the first prototype). They had no authority when the system was powered, but they did support the static weight (roughly), which allowed actuator sizes (& hence power consumption) to be reduced. They also provided a relatively "soft" failure mode. I drove a de-pressurized active vehicle for over 100 miles at motorway speeds on one wet & dark Sunday evening.

Edited by DaveW, 24 August 2010 - 19:41.

[Spoofski]

Sadly it would not be allowed in F1 but the development of radar for cars makes one wonder if a combination of specialised forward radar to map the terrain and active would be interesting.

I believe one or two F1 teams briefly investigated this in a collaboration with the FIA's safety commission about 10 years ago, around the time it became available on high-end MBs. It was intended as a safety-aid but I think it fizzled once it became clear that race-cars are nearly always about to crash into one-another and that discriminating between an accident about-to-happen and a normal nose-to-tail braking event would take a load of development.

[maxay1]

Very interesting stuff, particularly your thought experiment Dave....Thanks for the contributions everyone.

Wil

[DaveW]

Chassis roll (velocity, not angle?) may be an important cue for the driver at larger lateral accelerations, but in the absence of it, the driver is still subject to inertial reaction, I suppose. This should give some clue to the driver, as well as any o/s or u/s characteristic built into the control algorithm.

That thought bothered us too (not just me) before the event. With the benefit of 100% hindsight, I can state that it is not the case "obviously". There are often very few cues available for assessing roll angle in a modern vehicle. Also, how does a driver control a vehicle in wet conditions (when mu can be <10% of dry values). We found that allowing roll angle/gn actually interfered with control in the sense that it was accompanied by transients that could mask what the vehicle was doing. That was the case whether the vehicle rolled into or out of a turn. I think the absence of roll transients when manoeuvring was the feature that "real" drivers liked best. "Look how easy it is to control" was an Elio de Angelis comment whilst driving sideways, one handed, with armco in (uncomfortably) close proximity. Personally, I was busy hoping that a circuit board wouldn't drop out....

Lack of pitch angle with longitudinal acceleration did not appear to be a problem either, except that Senna liked some aero over-steer on turn in & mechanical under-steer on corner exit (at least, that was what he ended up with). I implemented the latter by moving balance to under-steer proportional to longitudinal acceleration (only when +ve). It was the only change I implemented at the track untested &, inevitably, I got the sign wrong. Senna ignored the "in" board for several laps before eventually stopping with a big smile & saying "I don't think that was what you intended". I still have a vivid recollection of him exiting Woodcote with a massive sideslip & full opposite lock (& still modulating his throttle). So much for worries that we might cause a driver to lose control....

Working with Ayrton Senna for three months or so was interesting & enduringly instructive. I think I really annoyed him only a couple of times. Once when I set the gain of the lateral balance algorithm too high (this effectively used steering input as a trajectory demand), & when I tested the "tidying up" integrators by setting simulated spring stiffnesses to zero. He really didn't like being able to sense that he was not in sole control of the vehicle.

Personally, I have only two regrets about the '99T project. The first was the we didn't understand the importance of (& hence did nothing about) driver "feel". I think that was behind Senna's complaint (comment, really) that "it always feels as though it is about to let go, but somehow doesn't". The second was that I think we were rather too successful in "hiding" basic vehicle deficiencies which, inevitably, became exposed at the true vehicle limit. This almost certainly cost Senna at least one race win, when he took to the gravel after overtaking a back marker. He said afterwards that "the system put me into the gravel, but then got me out again." That incident presented Williams GP with their first active win, I think.

Edited by DaveW, 25 August 2010 - 09:18.

[StressedDave]

@DaveW

Did you ever fiddle with the idea of variable rate damping on the road/race cars? We had a very brief play with a simple two-stage algorithm on the Scorpion, but never got beyond a first test, which proved inconclusive compared with optimised settings of the normal modal stiffness and damping. I always wondered, especially given the actuator travel we had to play with, whether it would have improved RMQ further if we could have had a 'blow-off' equivalent in the code. I guess the same thing may have been possible for kerb impacts.

[DaveW]

@DaveW

Did you ever fiddle with the idea of variable rate damping on the road/race cars? We had a very brief play with a simple two-stage algorithm on the Scorpion, but never got beyond a first test, which proved inconclusive compared with optimised settings of the normal modal stiffness and damping.

The quick answer to your question is no.

I didn't have much involvement with the Scorpion project, other than to suggest a track tensioning algorithm ahead of hardware being available, so I would be guessing about the details of the final control algorithms. However, for a more "normal" application, instantaneous actuator velocities were controlled largely by three elements. The simulated spring/damper velocities were added to the hub velocity feed forward signals, & the per-corner composite signals were modulated by a stroke-limiting algorithm that adjusted overall loop gain on a per-corner basis. The spring/damper component was complicated by the fact that modal (rather than "corner") springs & dampers were modelled. The hub velocity feed forward was calculated per-corner & I described this in principle in an earlier post. The stroke-limiting algorithm was suspiciously similar to something patented years later by the Lord Corp, & so can be found by searching on-line.

So, how to implement blow-off, bearing in mind that hydraulic actuators have to be driven everywhere? I think I would increase the hub velocity forward loop gain, triggered by a threshold corner vertical load (or, more accurately, the component of vertical load that was caused by road inputs). The limiting gain would be no additional force transmitted to the sprung mass, but the feasibility of achieving that would have to investigated, I think, as would managing the transition between the two conditions.

[StressedDave]

I'm fairly sure the algorithms for the Scorpion and the road cars were fairly similar, albeit minus the 'handling' algorithms which would have made little sense given the different steering system... I don't think the variable damping algorithm work in line with your thoughts, which might potentially explain the inconclusive test!

Our former colonials funded the track tensioning part of the programme, and as a result, a couple of the reports of the project are freely available online at:

http://www.dtic.mil/...mp;AD=ADA252824 &
http://www.dtic.mil/...mp;AD=ADA282723

For those who don't mind wading through pages of parameter files...

[DaveW]

For those who don't mind wading through pages of parameter files...

Thanks for that. Fascinating stuff....