Computing derivative using "for" loop--why am I getting this error?

Below is my code (note that in reality the vectors are much longer, but I simplified them to 5 entries of fake data...you get the picture). THIS code works, but when I use the actual data (each parameter is a 1x5036 matrix, as opposed to the 1x5 matrices that I present here for simplicity), I get a constant value of 0.75 for VehicleAccel for all t. Obviously, this is not the true value, as the data was for a real-world drive I did in a car around town (I have attached my actual m-file with the data). What is the problem?? Thanks in advance!

t = [1 2 3 4 5]; % sec

FLwhlRotation = [6.8 6.9 7.0 7.1 7.2]; % rad/s

FRwhlRotation = [6.8 6.9 7.0 7.1 7.2]; % rad/s

RLwhlRotation = [6.8 6.9 7.0 7.1 7.2]; % rad/s

RRwhlRotation = [6.8 6.9 7.0 7.1 7.2]; % rad/s

AvgWhlRotation = (FLwhlRotation + FRwhlRotation + RLwhlRotation + RRwhlRotation)./4; % rad/s

AvgWhlSpeed = AvgWhlRotation.*0.3683; % m/s (R = 0.3683 m)

VehicleSpeedMPH = AvgWhlSpeed.*2.2369; % mi/hr

VehicleSpeedKPH = AvgWhlSpeed.*3.6; % km/hr

% Instantaneous Acceleration (m/s^2)

VehicleAccel=0;

for i=1:4   
    VehicleAccel = (AvgWhlSpeed(i+1)-AvgWhlSpeed(i))./(t(i+1)-t(i));

end

%Plots

plot(t,AvgWhlSpeed,'b',t,VehicleAccel,'r')

legend('Speed (m/s)','Acceleration (m/s^2)')

xlabel('Time (s)')

[dpb]

Idea seems ok, but the attached file is worthless -- cut it down to a reasonable size and formatted text file that can read in...it's wrapped as shown and illegible.

I've revamped your code in "the Matlab way" using arrays and builtin functions to do the work...

t = [1 2 3 4 5].'; % sec
whlRot = [6.8 6.9 7.0 7.1 7.2; ...
        6.8 6.9 7.0 7.1 7.2; ...
        6.8 6.9 7.0 7.1 7.2; ...
        6.8 6.9 7.0 7.1 7.2].';
AvgWhlRot= mean(whlRot,2);
AvgWhlSpd = AvgWhlRot.*0.3683; % m/s (R = 0.3683 m)
VehMPH = AvgWhlSpd.*2.2369; % mi/hr
VehKPH = AvgWhlSpd.*3.6; % km/hr
% Instantaneous Acceleration (m/s^2)
VehicleAccel = diff(AvgWhlSpd)./diff(t);

I stored the data in column array; you can arbitrarily assign the wheel locations to column 1-4 as you choose to collect the data. As above it would be LF, RF, LR, RR

[Kelsey]

When I run that, I get an error message saying that it couldn't plot because the vectors aren't the same lengths (i.e., VehicleAccel is 1x5035, while t is 1x5036). Any idea what the issue is?

[dpb]

Of course...when you take the first differences there's one less value left than you start with. You don't show how you actually did the plot shown but your looping solution has the same issue; your initial code shows

for i=1:4
  accel(i)=(spd(i+1)-spd(i))./(t(i+1)-t(i));

which leaves you with four accel values calculated from a time vector of length five. I presume you probably augmented the accel vector w/ a leading zero.

The vectorized solution would be

VehicleAccel = [0 diff(AvgWhlSpd)./diff(t)];

Answer 1

[Star Strider]

Change this line to create VehicleAccel as a vector and see if that improves things:

VehicleAccel(i) = (AvgWhlSpeed(i+1)-AvgWhlSpeed(i))./(t(i+1)-t(i));

Also, it’s best not to use ‘i’ and ‘j’ as loop indices. MATLAB uses them for its imaginary operators, and could cause confusion if you use them also as variables.

[Kelsey]

Thanks!! That did it, I think...how does this plot look?

Would you say the acceleration appears to accurately represent the derivative of the speed?

[Star Strider]

My pleasure!

I would. The derivative should be slightly lower than baseline for the areas where the speed is linearly decreasing, and it seems to be. I opened a larger version in a new tab in Firefox, and it looks good. The derivatives of the parabolas look appropriately linear.